' W4Y+!.414 W L WWW .VI .1 Without . . . . UNIVERSITY OF CALIFORNIA. SAN DIEGO UC SAN DIEGO LIBRARY . w www www.www ww. . . W maman wwwwww . www .. . 3 1822 04429 6630 Dec 1989 OP1 Offsite STEMS GROUP TEC (Annex-JO NOTE NO. 216 rnals) QC 974.5 . T43 no. 216 HORIZON SCANNING IMAGER (PRELIMINARY FIELD TEST PLAN) R. W. Johnson UNIVERSITY OF CALIFORNIA SAN DIEGO The material contained in this note is to be considered proprietary in nature and is not authorized for distribution without the prior consent of the Marine Physical Laboratory and the Air Force Geophysics Laboratory RSITY CALI Contract Monitor, Dr. H. A. Brown Atmospheric Sciences Division VINTI with *!!!!! 11P ORNIA LLIT 1868, Prepared for The Geophysics Laboratory, Air Force Systems Command United States Air Force, Hanscom AFB, Massachusetts 01731 under contract NO. F19628-88-C-0154 SCRIPPS INSTITUTION MARINE PHYSICAL LAB San Diego, CA 92152-6400 OF este comanda si menina tor n in maniera ca.. i dans le is in den einen wei mit van een ander manier in diens with an mani na mwana walio ulit . de mise ... in primeren .. malainen kuin vastaano ni kama menerima suaminimin min standa nline tool nam nin naman ini mai mari mw mi n i nini kino indikana kami man seine ei OCEANOGRAPHY " . UNIVERSITY OF CALIFORNIA, SAN DIEGO me :* .. SAW .. * ... www . www REWITA www. CU w *.**** .. bewer wwwwwww.... ... .www .. ZA Wow...... ...... ** 3 1822 04429 6630 ***** * *** ** **** **** SI *vw v v w ?********* TABLE OF CONTENTS **...wwwwwwww w .** * ************ *****.. ******** * * *** ******** List of Tables and Illustrations WW *** * ******www. m 1.0 n S . e v i nc NNM Yalanim.. ..... ............ ...... 2.0 ............ .mmmmmmmmmmmmmmmmmmmmmmmmmmmmmmm...mmm...mmm. Introduction .. 1.1 Horizon Scanning Imager (HSI) Description ..... 1.2 Whole Sky Imager (WSI) Description : 1.3 Composite HSI/WSI System Description ....... Major Test Plan Elements ....... 2.1 Test Site Selection Criteria.......... 2.2 Data Collection & Reduction Procedures ........ 2.2.1 HSI Data Collection ......... 2.2.2 HSI Data Reduction Procedures. 2.2.3 Determination of Prevailing Visibility 2.2.4 WSI Data Collection ......... 2.2.5 WSI Data Reduction Procedures .... 2.2.6 Determination of Cloud Cover .... 2.3 Ground Truth Considerations. 2.3.1 Daytime Visibility ......... 2.3.1a Target Selection. .. 2.3.16 Inherent Contrast.. 2.3.1c Threshold Contrast. 2.3.1d Overcast sky conditions................. 2.3.le Partly cloudy horizon sky conditions.... 2.3.2 Daytime Cloud Cover ......... 2.3.3 Nighttime Visibility Determination 2.3.4 Nighttime Cloud Discrimination.. 2.4 Analysis Considerations ........ 2.4.1 Hardware Performance Analysis ..... 2.4.2 Algorithm Performance Analysis. : : : : : : : : : : : : 3.0 .......... Specific Test Sequences ........ 3.1 Test Site Selection Requirements........ 3.2 HSI/WSI Hardware Performance Testing 3.3 HSI/WSI Algorithm Performance Testing ... 3.4 HSI/WSI Nighttime Modification. 4.0 System Operational Considerations... 4.1 Set-up Requirements for Physical Stability. 4.2 Environmental Protection Devices ..... 4.3 Manual Mode/Automatic Mode Procedures ... 4.4 Data Archival & Recovery ....... Summary Comments 5.0 6.0 References and Bibliography LIST OF ILLUSTRATIONS Fig. # Title Page 1-1 Automatic Visibility System, as-built schematic ........ Tuy Automatic Horizon Scanning Imager System, as built exterior sensor Relative Spectral Responses . 1-4 Automatic Whole Sky Imager System, as-built schematic ........... Automatic Whole Sky Imager System, as built exterior sensor. ... 1-6 Composite HSI/WSI Accessory Control Panel........... 1-7 2-1 2-2 Composite HSI/WSI As-built Field Installation ....... Typical HSI Imagery. Menu Driven Display (automatic).. WSI Imagery, 512 x 512 with 33V x 33H Overlay . Single Wavelength Discriminators, Caveats. . 2-3 2-4 SI 2-5 WSI Basic Image Processing Flow Chart .......... 2-6 Relative Visual Range vs Target/Background Ratio ......... wwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwww... 2-7 2-9 ............... V/r, Sensitivity to Inherent Contrast. V/r, Sensitivity to Threshold Contrast . Visibility Error From Horizon Cloud, ys Cloud Brightness. Visibility Error From Horizon Cloud, ys Cid/Tgt Range WSI Basic Imagery, 650 nm ........... WSI Derived CLD/NO CLD image ......... 2-10 .................. 2-11 2-12 Table. # Title 1.1 Auto-Vis Horizon Scanning Imager (HSI), as-built Sept. 1988.... -iv- FIELD TEST PLAN VISIBILITY/CLOUD COVER SENSOR SYSTEM REF: CONTRACT F19628-88-C-0154 1.0 INTRODUCTION This test plan is proposed as a guide to the orderly and timely evaluation of an automated, all-weather image oriented system for the determination of local sector and prevailing visibilities, with their concurrent measurements of total cloud cover. model 2710 solid state video camera. This small, all solid state monochrome camera is well suited to this horizon scanning task. The CID sensor has a 512V x 776H pixel array yielding excellent angular resolutions particularly using the 100mm efl optical system which yields an approximate field of view of 5.4 degrees. This plan addresses several procedural facets including a) site location, b) data collection and reduction techniques, c) ground truth, d) analysis techniques, and e) recommendations. Most features of the automatic system to be tested are described in several earlier documents, prepared by the Optical Systems Group of the Marine Physical Laboratory, which are listed in the attached reference and bibliography list. The historical development and the as-built field configurations of the Horizon Scanning Imager (HSI) and its sister sub-system the Whole Sky Imager (WSI) are described in GL-TR-89- 0061, "Automated Visibility and Cloud Cover Measurements With a Solid-State Imaging System" from which much of the following descriptive material deemed useful in defining the test procedures has been abstracted. 1.1 Horizon Scanning Imager (HSI) Description The associated sub-assemblies most important to the reproducibility of the systems suite of selected target scenes are the Precision Rotary Table itself and the Microstepper Motor Drive. These two components, now co-located in the external housing, were originally isolated with the ustepper motor drive located within the accessory control panel (ACP). A variety of cabling and signal quality problems in the prototype configuration were resolved by this relocation and no further operational peculiarities have developed. The interior control console contains a computer configuration illustrated in Fig. 1-1 and Table 1.1. The majority of the systems primary control and archival processes are enclosed within the AT class micro computer. They include the video frame grabber which accommodates a 1024x1024 image memory, the 80286 based CPU card, and the 8mm cartridge tape system whose 2.2 Gigabyte storage capacity enables an extended automatic duty cycle for the fully automatic system. The Optical Systems Group at the Marine Physical Laboratory has produced a compact and reliable system for the automatic acquisition and archival of local horizon imagery specifically tailored to the determination and assessment of daytime sector visibilities, and subsequently their spatial and temporal variabilities. The as-built configuration for the system currently in operation at the Marine Physical Laboratory in San Diego is illustrated schematically in Fig. 1-1 and photographically in Fig. 1-2. The hardware listing appropriate to the illustrated configuration is given in Table 1.1. The use of a dedicated video system for the automatic determination of local sector visibilities offers the combined capabilities of solid state imaging technology and fast, efficient micro-computer processing. In the ratio oriented techniques associated with frame by frame assessment of the video imagery, the geometric stability of the silicon array with its exactly reproducible pixel placement is an outstanding advantage in assuring precise target/background definitions. Similarly advantageous to frame-by-frame analysis is the auto-iris characteristic of maintaining a video lines average detected radiance at approximately mid-scale on the detectors radiant flux versus analog voltage output curve. Due to the imaging chips reliably linear output in this mid-range region, highly precise ratios of target to background pixel radiances are readily obtainable. The exterior sensor assembly is a relatively compact, weatherproof housing which contains several inter related components and sub-systems. The detector component is the CIDTEC (formerly General Electric) Table 5t. Auto-VIS Horizon Scanning Imager (IISI) As-Built September 1988 AVC PWA DISTA RECIAC FAN - 15VDC CAM. PWA Exterior Sensor Assembly AUTORISE VIDEOEXTENDER Codi COLOR MONITOA SONY MODEL PVM - 12710 VIDEO CAMERA CID 776 ACCESSORY CONTROL PANEL MOTOR DRIVE MANUAL OVERAIDE 1. WXP Ilousing. E03-3900 2. CID TEC, 2710 Vidco Camera 3. Dacdal, 208018, Precision Rotary Table 1. Dacdal, MD23. J.Stepper Motor Drive 5. Cosmicar, C5018ES A Auto-Iris Lens 6. Cosmicar, Ex-2, Video Extender 1. Recirculation Fan 8. 15VDC Caincra Power Supply COMPUTER TMI 2001 MODEL 2001A PRECISION ROTARY TABLE DAEDAL INC. MOOEL 20801 P IMAGE PROCESSING SUB-SYSTEM ITI FG100. AT ARCHIVAL 3 VO SUB-SYSTEM DAEDAL MOOEL MD-223 USTEPPER MOTOR DRIVE 640 kbyte RAM 360 kbyte FLOPPY 65 Mbyte H.D. Interlor Control Assemhly UUU DAEDAL MC5000 INDEXING MOTOA CONTROL BOARD WEATHER PROOF ENCLOSURE MODEL E03.3900 Optional External VO Unit EXABYTE VIDEO COAX STOWED KEYBOARD 115VAC ACCESSORY POWER COMPUTEA TO MD-23 (DIR. $ STEP) COMPUTER TO TABLE (HOME A LIMIT SWX) FIG. 8 AUTOMATIC VISIBILITY SYSTEM HARDWARE LAYOUT - 8 DEC '88 AS-BUILT 1. TMI, 2001A, AT Compatible Computer A. TMI, B286-IM-AT CPU Board h. TMI, 80287-8, Math Co processer c. TMI, BVC-1, Color Video Card d. TMI, KBIAT, Keyboard e. Scagaic, ST-277R, 65MB Ilard Disc f. Western Digital, HICRA2, Disc Controller 8. Exabyle, EX-8200, 2.2GB Tape Streamer h. Adv. Storage Concepts, ASC-88, SCSI Host i. ICS Computer Products, DIO-96, TTLJO Ports j. (II. FG-100-1024-U-AT-A. Vidco Frame Grabber k. Compumolor, MC5000 Controller with PC-21 Indexer 2. Sony, PVM 12710. Color Monitor 3. MPL, E03-3000, Accessory Control Panel INTERIOR CONTROL CONSOLE 1-1 Automatic Visibility System, as-built schematic ... 1-2 Automatic Horizon Scanning İmager System, as built exterior sensor ... TA we The HSI camera system described in this note may readily be calibrated against standards of radiant intensity traceable to N.B.S. using standard radiometric procedures in association with optical calibration facilities established at the Marine Physical Laboratory. Several specialized facilities and fixtures are of course necessary to implement the calibrations specified for these camera systems. A three meter optical bench in conjunction with a selection of standard lamps and calibration targets can create the various radiance fields necessary to characterize the camera's performance in each of the following regions of interest. a) Radiometric linearity - radiative flux vs byte value output. angular resolutions even with the super wide angle (1749) fisheye optical system. The charge injection device has inherently low blooming characteristics resulting in no smear and no lag highlight handling. Background fixed pattern rejection circuitry contributes to excellent low illumination image quality, and the cameras RS-170 compatible video output makes interfacing with peripheral display and processing devices straightforward. The three associated sub-assemblies, the solar occultor, the remote controlled iris and the remote controlled optical filter assembly have been fabricated at the Marine Physical Laboratory. These peripheral components provide stray light and flux control to ensure optimum image quality over a broad range of illumination levels, and also to provide for the selection of the task specific spectral filters. The optical filter assembly contains two independently controlled four position filter wheels, one contains four neutral density filters for flux level control, and the other contains four spectral filters for spectral pass band selection. All three sub-assemblies are controllable through the accessory control panel, located within the interior controller installation. Each sub-assembly can be operated under either manual or computer control. b) Absolute radiative response - absolute spectral flux input vs byte value output. (Not normally applicable to HSI systems) c) Relative radiance shifts - systematic changes in flux/byte relationships as a function of f-stop and neutral density changes. d) Geometric mapping - pixel element vs object space geometry as diagnosed with Precision Calibration Target. The majority of the systems primary control and archival processes are enclosed within the AT class microcomputer. They include the video frame grabber which accommodates a 1024x1024 image memory, the 80286 based CPU card, and the 8mm cartridge tape system whose 2.2 Gigabyte storage capacity makes possible the seven day duty cycle of the overall camera system. Calibrations of the sort outlined above for the HSI system, are conducted in the green spectral response band illustrated in Fig. 1-3. 1.2 Whole Sky Imager (WSI) Description The Optical Systems Group at the Marine Physical Laboratory also has produced a compact and reliable system for the automatic acquisition and archival of whole sky imagery specifically tailored to the assessment of daytime cloud distributions and subsequently their spatial and temporal variations. The as-built configuration for the whole sky systems currently (Sept 88) being fielded is illustrated in Figs. 1-4 and 1-5. A brief discussion of the most significant features of this E/O Camera V system is contained in the following paragraphs. The exterior sensor assembly, shown schematically in Fig. 1-4 and photographically in Fig. 1-5, is a relatively compact, weatherproof housing which contains four independent but inter-related sub-systems. The detector component is the CIDTEC (formerly General Electric) model 2710 solid state video camera. This small, solid state monochrome camera is well suited to this whole sky imaging application. The CID sensor has a 512Vx776H pixel array yielding excellent Whereas many useful algorithms for the determination of atmospheric properties can be devised to require only the input of the relative values of radiant flux fields, it is generally true that far more redundant and reliable methodologies are available when absolute values of radiance are available. Thus, to enable an optimum selection of techniques for analytic applications, the camera systems described in this note are all calibrated against standards of radiant intensity traceable to N.B.S. using standard radiometric procedures in association with optical calibration facilities established at the Marine Physical Laboratory. 1.3 Composite HSI/WSI System Description . The prototype visibility and cloud detection systems, while very similar in nature and using many common hardware sub-assemblies, are two separate hard- ware/software entities which need not necessarily be RELATIVE SPECTRAL RESPONSE Oriel Broadband Interference Filters with GE2505 Camera Image Acquisition & Analysis System Hardware Block Diagram E/O System 5 1.0 0.94 Peak Bandwidth - À @ 0.5 Peak 456 56.7 550 54.7 654 71.1 0.8 blue green red SONY PVM 1271Q MONITOR Relative Response, S(A)T(A) o v in ΤΥΤΤΤΤΤΤΤ ΤΤΤΤΤΤ w i GE 2710 SOLID STATE VIDEO CAMERA TMI COMPUTER (IBM/AT CLONE) 02 AUTOMATIC EQUATORIAL SOLAR OCCULTOR ASSY. VIDEO IMAGE PROCESSING SUB-SYSTEM (ITI FG 100) ARCHIVAL 110 SUB-SYSTEM (SEAGATE 65 M byte H.D.) 0.0 400 REMOTE CONTROLLED IRIS ASSY. 450 500 700 750 800 EXABYTE EXB - 8200 2.2 G byte 8mm CARTRIDGE TAPE SYSTEM 550 600 650 Wavelength (nm) ANALOG ACCESSORY CONTROL PANEL REMOTE CONTROLLED OPTICAL FILTER ASSY. 1-3 Relative Relative Spectral Responses ..... STOWED KEYBOARD EXTERIOR SENSOR INSTALLATION INTERIOR CONTROLLER INSTALLATION 1-4 Automatic Whole Sky Imager System, as-built schematic. 350 1-5 Automatic Whole Sky Imager System, as built exterior sensor. ..... QA ..... ..... ... ..!!!!!.... ...... . ........... . . ......... .......... . .. . .... .. ....... . two camera sub-systems. These modifications do not affect the fundamental design or operation of the HSI and WSI sub-systems. Their primary function is to provide adequate provision for the environmental control sub-systems. An as-built illustration showing the composite system which is operational on the MPL rooftop data site is shown in Fig. 1-7. 2.0 MAJOR TEST PLAN ELEMENTS ............ .. . .. .... .. ............ ... .. ........ ..... . .."" """ . integrated into a single unit. Each may stand alone and function independently. This separation was intentional during the early stages of development such that each could be optimized in its task performance attributes without being compromised by conflicting packaging constraints. However, with several versions of each system successfully operational it became appropriate to initiate a merging of the two devices into a single entity. As might be expected, a complete re-design and refabrication of two optical-mechanical housings was deemed overly expensive at this stage of development and a less expensive yet more instructive procedure was chosen for implementation. Two relatively straight forward modifications were performed to achieve the merger, one in hardware and one in software. The salient points of each are discussed briefly in the following paragraphs. The hardware modification was the simplest. It involved designing and building a new Accessory Control Panel (ACP) which contained all of the features originally contained in the two original stand alone units. The electrical interfacing was not particularly difficult due to an earlier re-location of the HSI Micro Stepper Motor Drive sub-assembly into the system's external weather proof enclosure. The new composite ACP is illustrated in Fig. 1-6. Part of the overall research effort to provide and evaluate an automated weather sensor system, its related techniques for the observation, analysis and display of sector and prevailing visibility, and cloud cover conditions for continuous day/night, all weather operations, has been the development of this compre- hensive Field Test Plan. In the following sections it addresses, (a) site selection, (b) data collection and reduction techniques, (c) ground truth, (d) analysis techniques, and (e) recommendations. Whereas some of the test sequences listed in Section 3.0 may imply data quantities beyond the minimum acceptable for engineering validation, they have been proposed for inclusion in the interests of optimization of user confidence levels. 2.1 Test Site Selection Criteria The software modifications required for implement- ing the dual system were more severe, but not unreasonable. Had the control computer been running a 80386 based CPU card, which would enable more realistic multi-tasking instead of the existing 80286 based card, some simplifications would surely have been available, however the upgrade was not made at this time. As additional procedural sophistications become appropriate, the changeover can be made at any time required. Under the control of the composite system software, the HSI/WSI retains its ability to run under either manual or full automatic mode. It will display its output products of sector and prevailing visibilities plus total cloud cover on the CRT, or output to printer. It will also, at the operators discretion, archive both original imagery and derived numerical products to the EXABYTE tape sub-system for later retrieval and analysis. Several additional mechanical modifications were made to the weather proof enclosures, and field mounting structure to enable physical co-location of the In establishing criteria for ensuring the selection of a suitable test site, one should recall that daytime visibility determination, either instrumentally or by a human observer, ultimately involves the resolution of the distance from pre-selected targets that the apparent contrast, Cr, reduces to some minimum (threshold) value needed for detection. The threshold contrast depends on a number of factors including the visual acuity of the observer as well as the angular subtense of the target, its shape and its location with respect to background features. Human estimates of visibility result from adaptive integration by the eye in time and space of all scene features. The World Meteorological Organization (WMO) and the Federal Meteorological Handbook (FMH) recommend that daytime visibility targets be black or very dark targets viewed against the background sky. However, in order to achieve adequate areal coverage, objects selected for visibility markers for routine meteorological observations often are less than ideal both in angular size and inherent optical properties. If one assesses the scene captured in Fig. 2-1 it is clear that a solid state imaging system, in this case the HSI, can provide highly accurate and readily updateable measurements of the apparent contrast of all identifiable EO CAMERA VISIBILITY SYSTEM ACCESSORY CONTROL PANEL TEMPERATURE LOCAL OCM CCW CW WSI VIS TE REMOTE ROTARY TABLE SPECIAL NEUTRAL DEN LOCAL LOCKOUT SWITCH DISABLE dD LOCAL POWER OPEN CLOSE IOCN TRGO HENTE VA GOI POTE CHANGER APERTURE 1-6 Composite HSI/WSI Accessory Control Panel... 1-7 Composite HSI/WSI As-built Field Installation .... 3A objects within its field of view. Therefore it is incumbent upon the selected test site to enable the best possible match between the camera systems inherent capabilities and the maximum availability of targets, optimized for visual estimates in accordance with both the FMH and the appropriate underlying psychophysi- cal relationships discussed in the attached bibliography. Additional comment in this regard is contained in Section 2.3.1. 2.2 Data Collection and Reduction Procedures In its current field configuration, the Horizon Scanning Imager sweeps its field of view through the entire 360° local horizon in approximately one minute. The sweep is momentarily interrupted at up to 32 pre- selected azimuths for target assessment and line of sight visibility determinations. At each stop, up to six different targets of opportunity may be selected for use in ratioing against the local horizon radiance. The field of view of the camera system is approximately 5.4 degrees which produces an image at each azimuth stop similar to that shown in Fig. 2-1. Every 10 minutes, or as instructed by its user, the composite HSI/WSI system interrupts the visibility data sequence and jumps to a cloud cover determination such that both sector and prevailing visibilities plus intermittent cloud cover data are consecutively cycled throughout any pre-specified data interval. 2.2.1 HSI Data Collection visibility that are consistent regardless of time and location. The degree of visual modeling required to provide, for example, the visual range along specific pre-selected slant paths depends upon the accuracy needed to satisfy the individual user requirements. In the HSI, we limit attention to the more general problem of determining prevailing visibility, which is defined by Douglas and Booker (1977) as the greatest visibility equal or excluded through at least half of the horizon circle which need not be continuous, i.e. the median visibility around the horizon circle. The E/O camera system provides a continuously updated representation of the detailed radiance scene for the discrimination of prevailing visibility and its sector variations. The objective is to exploit the measured apparent radiance contrast for all suitable objects in the scene, adjusting the resultant visibility estimates as necessary to compensate for the non-standard nature of individual targets. Several computer displays of the measured and derived data used by the machine in its computations have been devised to assist the operator during initial set-up sequences, and later for either archival or quality control procedures. The typical data scene illustrated in Fig. 2-1 contains a variety of targets of opportunity, those outlined having been selected to determine the visibility for this specific scene. A currently used display is presented in Fig. 2-2. This display is the output from a fully automatic sequence, in which the operator's set-up parameters have been pre-selected and frozen into the computational algorithm. At this point, the camera system goes through its automatic horizon sweep routine, stopping at pre-designated azimuths and calculating the visibility for each scene. Each designated target within the scene is used to make an independent estimate of visibility, and the result displayed as illustrated. Computed values of visibility are tagged to indicate whether or not the target used is appropriate for optimum computational reliability. For targets with ranges too near, the computed visibility is tagged ">", i.e. greater than, similarly for targets deemed too far, i.e. undetectable, the computed visibility is tagged "<", i.e. less than. im III As noted earlier, in its automatic data collection mode the HSI system acquires a set of images similar to that shown in Fig. 2-1 every minute. In each image a series of targets at known ranges have been pre-selected by the operator in an initial set-up sequence. Using these images the machine is programmed to simulate the human observer's assessment of the scene in a manner consistent with the appropriate definitions and methodologies discussed in the attached references. These assessments are displayed to the user's computer screen, and at his discretion automatically archived to tape for later retrieval and analysis. 2.2.2 HSI Data Reduction Procedures STOP (2-106) RANGE ((MILES) THRESHOLD CONTRASH INHERENT APPARENT HORIZON CONTRAST CONTRAST BRIGHTNESS VISIBILITY (MILES) TARGET 1 25.00 .030 .800 <.030 161. < 25.0 TARGET 2 15.00 .030 .800 <.030 161. < 15.0 TARGET 3 6.00 .030 .800 .185 161. TARGET 4 3.00 3.00 .030 .030 .300 .800 349 161. 161. 11.9 The goal is to develop a completely automated system to provide objective quantitative estimates of Fig. 2-2 Menu driven display (automatic). 2-1 Typical HSI Imagery ..... # SI Oddelen W laconenden & 2-3 WSI Imagery, 512 x 512 with 33V x 33H Overlay .. HA ...... ... .. .......... . ...... . .......... .. ................ .. . ........ ............. . .. .. ............ .. v ies!!!... i v v .!!! W !W !!!! .!! ..... .. .... .... .... ...... . ...... .. ... period of two years. In accordance with this general time line, a basic set of four images may be captured during the first eight seconds of each minute. These four images, two red and two blue, are digitized by an 8 bit A/D converter and subsequently may be stored at either 512X512 pixel resolution or in a lower resolution sub-set of 33 rows and 33 columns. Slower grab rates are programmable if desired. For composite HSI/WSI operations the cloud imagery is currently acquired every 10 minutes and cloud cover estimates are made on line by the computer to provide near real-time numerical products for simultaneous display with the visibility determinations made by the HSI. ......... .. .. .. ........ . .. ...... ... . ..... .... . ........ .. ............................. .. .. ........ .. .. .. ... .. . ............. .................. It is evident that one must be careful in the interpretation of visibility observations where the human observer finds it necessary to depart from the guidelines and selects non-standard targets with backgrounds other than the adjacent horizon. Although digital imagery offers added opportunity to extract information with respect to the optical properties of the ambient atmosphere, considerable sophistication must be built into algorithms that diagnose effectively the detection range of many non-standard targets against variable backgrounds. On the other hand, the wealth of information generated by repetitive horizon scans with a digital imagery system enables straightforward determinations of prevailing visibility and visual range from a hierarchy of targets. The resultant measurements can be designed to conform in accuracy and representativeness with the highest standards of current practice. 2.2.3 Determination of Prevailing Visibility The preferred approach for instrumental determinations of visibility from radiometrically calibrated imagery follows directly from the discussion of the factors that govern image transmission in the atmosphere. First and foremost it is important, insofar as possible, to direct attention to the measured apparent contrast between the horizon sky and objects in the scene. It is desirable but not strictly necessary that the objects stand above the horizon as viewed from the ground observation point. The major consideration is that the extinction and directional scattering properties along the paths of sight to the object and horizon correspond closely. As discussed in GL-TR-89-0061 with the horizon sky as the background, the sky ground ratio is 1 and the systems operative equation becomes An example of the digitized resolutions appearing in the raw imagery is shown in Fig. 2-3. This monochromatic image displays both the full 512x512 resolution initially captured minute by minute in each of the four images per set, and as an overlay, the 33 rows and columns that comprise the archived one minute sub- sets. V/r=In() in C) Scene identification information is automatically imbedded into each image at the time of acquisition which specifies site, location, time, date, and accessory control settings. These data are decoded by the systems post-archival quality control programs and superimposed on all displayed imagery as seen in the upper left corner of Fig. 2-3. The black square area near the center of the image in Fig. 2-3 is an equatorially driven solar occultor to shadow the fisheye lens, thus preventing undesired stray light from entering the optical path. Each WSI field system is designed to run automatically and, if desired, without operator intervention for seven days, at which time the system shuts down, ejects its 8mm data cassette, and asks the field operator to insert a new tape into the recorder, which automatically restarts the system on a new seven day cycle. Alternative duty cycles for the composite HSI/WSI system are programmable in software. 2.2.5 WSI Data Reduction Procedures A variety of techniques for objective cloud/clear-sky discrimination with the solid-state imagery system are being investigated. Single image (monochromatic) analysis with simple brightness contrast and/or edge depiction techniques often provides excellent results. The cloud radiances are in general larger than the background clear-sky radiances, especially in the red portion of the visible spectrum. Although the (2.1) With careful selection of visibility targets on or near the horizon, Eq. (2.1) becomes the basic expression for the determination of visibility from the horizon scan imagery. Criteria for target selection and the specification of inherent contrast Co and threshold contrast e as input variables to Eq. (2.1) are reviewed in section 2.3. 2.2.4 WSI Data Collection Each WSI system has been designed to acquire automatically, whole sky imagery in two spectral bands, i.e. Blue @~450nm and Red @~ 650nm, at a rate of one set per minute for twelve hours a day over a ............ .... .. ...... ..... .... ............ ... . .. . . . . .. , . Well dolwww d doo o w i ..... qualitative brightness contrast is readily apparent in most images, obvious difficulties are involved in the development of objective depiction criteria using single wavelength image analysis. As summarized in the list of caveats in Fig. 2-4, the cloud and clear-sky radiances vary over broad limits, requiring complex analytic adjustments to the appropriate cloud/clear-sky threshold radiance values. Furthermore, dense and/or shadowed cloud radiances are comparable and sometimes become less than the corresponding background sky radiances. ..... ............. .. ........... .. . ...... . 1. Radiance Threshold Techniques a. Sky & Cloud Radiances Vary Over Large Excursions b. Threshold Radiances Require Continual Analytic Updates c. Dark or Shadowed Clouds Mimic Background Sky Radiances. cloud discrimination center primarily on paths of sight near the horizon, especially the upsun direction near sunrise and sunset, where the blue-red spectral ratio of the background clear-sky typically becomes less than one in hazy atmospheres. The problem also extends to the solar aureole region at higher solar elevation angles in dense haze conditions where little color contrast exists between the white aureole region and the existing clouds. 2.2.6 Determination of Cloud Cover A numerical determination of opaque, thin and total cloud cover can be made from each set of four basic images acquired by the WSI system. The necessary processing and computational steps are illustrated in Fig. 2-5. This stepwise conversion of the raw video imagery into automatically determined pixel by pixel specification of cloud or no cloud can, of course, be as sophisticated as the end use requires. The intent of this methodology is to impose strict quality and calibration controls on the raw data, while simultaneously reducing the number of images required for insertion into the optimized cloud/no cloud decision processes. As implied in the processing flow chart, it is inevitable that there will be a variety of specialized decision algorithms which must be applied to provide accurate cloud discriminations under the myriad of real world meteorological circumstances. 2. Edge Gradient Techniques a. Cloud Imbedded in Cloud Background Yields Redundant Boundaries b. Fails as Scene Approaches Uniform Overcast Fig. 2.4 Single wavelength discriminators: Caveats. On the other hand, experimental results clearly demonstrate the effectiveness of color contrast, as derived from multi-spectral imagery, for objective sky- cover analysis. Dedicated systems with the dual filter wheel option can easily be configured to acquire sequentially narrow-passband blue and red imagery with variable neutral density control. Objective cloud depiction algorithms based upon the ratio of the all-sky blue and red radiance fields provide good specification accuracy over a broad range of sky and visibility conditions. The blue/red spectral radiance ratios are characteristically near 1 for the white or grey cloud elements in contrast with significantly larger values for the background clear sky. Numerical examples demonstrating the efficacy of using blue-red radiance ratios for cloud discrimination can be readily generated using standard radiative transfer models with assumed levels of atmospheric aerosol loading. Calculations made with the Hering atmospheric radiance model FASCAT (AFGL-TR-84-0168), indicate the blue/red radiance ratio for cloud-free sky to be substantially elevated above the same ratio for those sky regions containing clouds. Sample calculations illustrating this characteristic are shown in GL-TR-89- 0061. Current techniques provide for either post archival, or on-line determinations in accordance with the illustrated sequences. Obviously for near realtime outputs, optimized decision algorithms must be pre- programmed in accordance with the user's specification. 2.3 Ground Truth Considerations 2.3.1 Daytime Visibility Under computer scan control, a digital image- acquisition system continuously maps the apparent radiance distribution and the relative apparent contrast of objects within the view of the observation point. The measurements are immediately applicable to the determination of the optical and meteorological properties of the ambient atmosphere. In the case of meteorological visibility, the objective is to extract numerical estimates commensurate with the human perception of visibility. In this regard, guidance given by the World Meteorological Organization (1971) to help ensure observation compatibility and representativeness prescribes: "Meteorological visibility by day is defined as the greatest distance at which a ... Related problem areas that are associated with the independent use of the blue-red ratio algorithm for BASIC IMAGERY CORRECTED IMAGERY COMPOSITE RATIO DELIVERABLE DATABASE CALIBRATION FUNCTIONS RADIANCE CONVERSIONS -- - CLOUD/NO-CLOUD DECISION ALGORITHMS BLUE IMAGE 512x512x8 CALIB. BLUE IMAGE RADIOMETRIC LINEARITY BLUE RED RATIOS RADIOMETRIC SENSITIVITY UP-SUN RED IMAGE 512 X 512 x 8 CALIB. RED IMAGE DERIVED PRODUCTS DOWN-SUN OPTICAL DISTORTIONS IMAGE RATIO COMPUTATIONS NEAR HORIZON SPATIAL DISTRIBUTIONS COMPOSITE BLUE/RED RATIO IMAGE OPTIMUM CLOUD/NO-CLOUD IMAGE FIELD OF VIEW DEFINITION TWILIGHT TEMPORAL DISTRIBUTIONS BLUE +N.D. IMAGE 512 X 512 x 8 CALIB. (BLUE) IMAGE DAWN SENSOR CHIP UNIFORMITY STATISTICAL PARAMETERS - AS REQD PIXEL SELECTIONS FOR OPTIMUMIZED COMPOSITE - (BLUEV(RED) RATIOS - REGISTRATION ADJUSTMENTS - - RED + N.D. IMAGE 512 X 512 x 8 CALIB. (RED) IMAGE - FLUX CONTROL | THRESHOLDS - - 2-5 WSI Basic Image Processing Flow Chart. Hintoiatu YUMIN LAKI TWO 01.liJIINTIM OY: WWII TSMMM 00: 11:1. NION Se .. ............ 2-11 WSI Basic Imagery, 650 nm. 2-12 WSI Derived CLD/NO CLD image GA ... ...... .. .. ............. ...... ..... ........ . .. ...........!!!!....... ................ ....... .................... .... .... .... .. . ........ detection, where the most consistent and representative calculations of visual range can be made. The important consideration is that the objects in a selected-area ensemble be located at approximately the same angle and as close in range as practicable to the limit of visibility. ....... .. ...... . . .... ....... .... . . .. ................. ......... ..... .... black object of suitable dimensions, situated near the ground, can be seen and recognized, when observed against a background of fog or sky." For both practical and theoretical reasons, the technique for the instrumental determination of visibility should conform with all elements of the definition. As discussed in the following paragraphs, strict adherence to these criteria for target/background selection results in a simple direct relationship for the numerical representation of visibility, V, through Eq. (2.1), greatly enhancing the interpretation of measurements made in different places at different times. 2.3.1a Target selection ....... ............... ................. ................. .................................... .......................................................................................... .... The prime requirement for reliable determination of prevailing visibility is an ample supply of suitable targets, well distributed in range and azimuth. On the one hand it is important to make use of all objects in the horizon scan imagery that may serve as visibility targets. On the other hand, because of their intrinsic properties, potential targets are not equal in a given situation as effective determinants of the prevailing visibility. The basic properties of individual targets must be predefined by the input variables to Eq. (2.1). In addition to the target distance from the observation point, the input variables are the inherent contrast of the target and the contrast threshold for detection. The effect on visual range determinations of using non-black targets is illustrated in Fig. 2.6. In this plot, the horizontal axis is conveniently marked in both inherent reflectance ratio and the equivalent inherent contrast. The vertical axis is marked as Relative Visual Range, which is defined as the ratio of the visual range one would determine observing a non-black target, to that one would determine if observing an ideal black target having the optimum inherent contrast of minus one. Stability in the input variables over the full range of atmospheric conditions is the major consideration in the selection of specific objects to be included in the accepted group of visibility targets. A discussion of the detailed elements involved in the determination of atmospheric visibility is contained in the Section 6 of GL-TR-89-0061. 2.3.1b Inherent contrast For visibility determination using Eq. (2.1), we must prescribe the relative contrast between the inherent radiance of the selected object and the adjacent horizon sky radiance. In the case of the ideal, non-reflecting, black target the relative contrast is always -1. (Note that the measured apparent contrast of a given target- background does not change sign with increasing path length so that the absolute value of contrast can be used for calculation purposes). However, the reflectivity of even natural dark targets is seldom zero so that careful attenuation must be given to the estimates of inherent contrast and the vulnerability of the target to fluctuations in Co due to changes in the directional distribution of lighting reaching the target from the sun, sky and surrounding terrain. The sensitivity of visibility determinations to the uncertainties in the estimates of the inherent contrast as an input variable is illustrated in Fig. 2-7. Trial calculations were made of the visual-range/target-range ratio as a function of measured apparent contrast for selected values of assumed inherent contrast. Shown here are the disparities in V/r associated with departures from an assumed inherent contrast of 0.8 for natural dark targets in the acquired imagery. The range of assumed values, -0.6 to -0.1, is representative of this class of targets. Note in particular that the visibility determinations are not sensitive to errors in the input values of Co when the object is near the maximum distance at which it can be seen so that the apparent contrast is close to the threshold detection contrast. However, the error sensitivity increases substantially for nearby targets in good visibility conditions when the measured apparent contrast is relatively large. Thus, the relative accuracy of visibility as determined from a given non-standard target is in part a function of the visibility itself. Most weight in a given situation should It is apparent that the relative visual range determination to be within 5% of the ideal, negative contrast can range between -1 and -0.86. However, for varying positive contrasts, the relative visual range changes quite rapidly, and the 5% variance from ideal is contained in a very narrow band between +1.9 and +2.2. Whenever practicable, as illustrated in Fig. 2-6, the selection of visibility targets should be adjusted for each scene in accordance with observed conditions. In the absence of ideal dark targets viewed against the background horizon sky, priority should be shifted as necessary to objects at distances close to the limit of INHERENT CONTRAST -.99 -.95 ..9 ..8..7 ..5 0 +1 +2 +4 +9 Www . . RELATIVE VISUAL RANGE - - - - ..... Target Darker Thon Background ...... Target Brighter than Background . 10 10 INHERENT TARGET TO BACKGROUND RATIO (RAR) 2-6 Relative Visual Range vs Target/Background Ratio .... . VISIBILITY I TARGET RANGE SENSITIVITY TO INHERENT CONTRAST VISIBILITY I TARGET RANGE .. SENSITIVITY TO THRESHOLD CONTRAST . viclco o CO = -0.6) TARGET RANGE A co ovo -1.0 ori o A VISIBILITY / TARGET RANGE W W THRESHOLD CONTRAST VISIBI 0.02 0.05 0.1 0.6 0.6 0.2 0.4 APPARENT CONTRAST V/r, Sensitivity to Inherent Contrast ...... V/r, Sensitivity to 0.2 0.4 APPARENT CONTRAST V/r, Sensitivity to Threshold Contrast ........ . . . . . . . . . . . . . . . 2-7 2-8 .. . .. MDMA -en va . .. vi.. . WAS IN T H E VISIBILITY ERROR FROM HORIZON CLOUD VS APPARENT CONTRAST AND CLD BRIGHTNESS VISIBILITY ERROR FROM HORIZON CLOUD VS APPARENT CONTRAST AND CLOUD RANGE 180- Co =-.85 CLD / TARGET RANGE = 2 160 - Co=.85 CLD / HORIZON SKY RADIANCE RATIO = 2 140 - BRIGHT CLOUD BRIGHT CLOUD RANGE TARGET RANGE 1.5 PERCENT ERROR IN VISIBILITY PERCENT ERROR IN VISIBILITY 204 N ** +# * W * * ** + . * *** . + DARK CLOUD N 5.2 0.4 0.6 . 0.6 MEASURED APPARENT CONTRAST Visibility Error From Horizon Cloud, ys Cloud Brightness. 2-10 7A 0.2 MEASURED APPARENT CONTRAST Visibility Error From Horizon Cloud, vs Cid/Tgt Range 2-9 be applied to the fraction of preselected targets where the measured contrast transmittance of the intervening path is relatively low. 2.3.1c Threshold contrast Experiments by Taylor (1964) have shown that the visual contrast threshold does not vary significantly with background luminance over the normal range of daylight conditions. However, the contrast threshold does vary markedly with the angular subtense of the target at the point of observation. With dwell times commensurate with normal visual search (1/3 sec), the minimum contrast for confident detection by the human eye is about .025 for objects with an angular size greater than 0.5 deg, as determined from the laboratory experiments. Based upon general field experience, Douglas and Booker (1977) recommended "---a recognition contrast threshold value of about 0.05 when measurements of transmittance are used for determining the visual range equivalent to that reported by meteorologists and the use of a value of e in the region of 0.035 to 0.04 for the detection contrast threshold under field conditions". The sensitivity of the visibility determinations to the assumed value of threshold contrast is illustrated in Figure 2-8. 2.3.1d Overcast sky conditions As in the case of a cloudless horizon sky, the apparent horizon radiance of an overcast sky tends to remain unchanged as the observer backs away from the target. Thus, Eq. (2.1) also holds for the overcast horizon case, and the resultant determinations of visibility are commensurate with the fundamental definition of meteorological visibility. However, the inherent contrast of dark objects with the overcast horizon is normally greater than the corresponding inherent contrast with the cloudless horizon, yielding an overestimate of the visibility unless proper adjustment is made in the input values of inherent contrast. 2.3.le Partly cloudy horizon sky conditions Horizon sky backgrounds with variable cloud conditions introduce additional uncertainty into the visibility determinations. Shown in Figures 2-9 and 2- 10 are the results of trial calculations illustrating the errors in V associated with the presence of isolated clouds on the horizon when a cloudless horizon is assumed for the calculations. Key factors are the distance of the cloud background relative to the target and the brightness of the cloud relative to the corresponding clear sky horizon radiance. Again we see that the resultant disparities are small when the contrast transmittance of the path of sight is small, regardless of the cloud position or brightness. The errors increase substantially in the case of targets having a range much less than the existing visual range. As shown in Figure 2-9, without an appropriate change in Co, the presence of a dark cloud results in progressively larger underestimates of visual range with increasing apparent contrast. The errors due to clouds much brighter than the corresponding clear sky horizon radiance are opposite in sign and more severe. The comparison in Fig. 2-9 gives results for horizon cloud radiances equal to twice and one-half the cloudless horizon radiance. The assumed inherent contrast with the clear-sky horizon is -.85 and the assumed cloud/target range ratio is 2. The sensitivity with respect to relative cloud distance is shown in Figure 2-10. A cloud/clear-sky brightness ratio of 2.0 and Co equal to -0.85 was assumed. For this bright cloud example, the error increases well beyond 50% for a cloud positioned at a range less than 1.5 times the target range when the apparent contrast is greater than 0.3. The error diminishes significantly for cloud ranges more than twice the target range, and of course for cloud brightness closer to that of the clear sky horizon brightness. 2.3.2 Daytime Cloud Cover Unlike the determinations of contrast transmittance over extended paths where a variety of contributing measurables and their impact upon the determination can be independently assessed, there are few data sources for independent assessment of the temporally fleeting state of the local cloud cover. For after-the-fact comparisons one has tedious access to hourly meteorological observations if there is an accessible reporting station in the vicinity, and to a lesser degree of suitability there are satellite images for regional assessments. For the most part, an automatic system operating at a random data site must provide adequate information to the user to enable stand alone validation of its performance without the luxury of an externally derived data set. The WSI system provides a method for achieving a reasonable level of self-testing by enabling the user to archive for subsequent comparison both an original high contrast whole sky image and also the corresponding machine derived cloud/no cloud decision image. An example of self test imagery is illustrated in Figs. 2-11 and 2-12. Fig. 2-11 is one of the "basic imagery" images acquired and archived every ten minutes by the WSI system. This high contrast whole sky image, in 9 Si : V isi i n i . *** * ** * * * *** ** *** ******* *** * * * system, the control computer will have to shift to transmissometer mode, using pre-selected fixed lights as radiance sources. Supplementary algorithm development will be necessary to allow the conversions relating atmospheric transmissivity, source intensity, illuminance threshold and visual range. An analysis of Allard's Law, its relation to the visual detection of fixed lights, and conversions to equivalent daytime/nighttime visual ranges is provided by Douglas & Booker, (1977). 2.3.4 Nighttime Cloud Discrimination * **** *** * ** * *** * ****** . . . . . . . . . . addition to being a basic data input to the processing cycle, also can be usefully employed as the bogey "sky truth" image against which the automatic determination of cloud/no cloud can be judged. The false colored image shown in Fig. 2-12 represents the results of a computerized pixel by pixel discrimination of whole sky cloudiness. It was derived by exercising the processing sequences illustrated in Fig. 2-5. To the degree that one might associate the visual interpretation of Fig. 2-11 with an observers estimate of cloud cover, then the machines assessment as shown in Fig. 2-12 can be objectively related to "truth". It is important to note that the WSI field of view is well documented, the solid angle of observation by each pixel within the image is fixed and known. Thus the fractional portion of the sky dome representing either cloud or clear areas is precisely and reproducibly determinable unlike the estimates of even well trained observers. . . . . . . . 2.3.3 Nighttime Visibility Determination Extending visibility determinations into the night- time regime provides a broad spectrum of potential techniques each with its share challenging research opportunities. For the most part, the choices are divided primarily into using either passive or active systems. In this context, we would restrict the definition of "active" to imply the use of an emitter as in a transmissometer light source, or an external probe as in a lidar. The term "passive" implies a staring or scanning system that acts only as a non-interrogating receiver of information from its surrounding environment. The detection and measurement of cloud field characteristics with the whole sky imager depends primarily upon the radiometric sensitivity of the camera system, and the spectral relationship between the background clear sky and the cloud elements superimposed against it. It has been well demonstrated that during the daylight hours, the requirements for both of these parameters are quite satisfactory, i.e. the camera can readily acquire high resolution imagery, and the blue/red sky radiance ratio clearly discriminates between clear blue sky and the non-blue cloud elements. It is reasonable to assume that if these two conditions were to be extended into the nighttime hours, then the cloud detection system would continue to function normally. It is our contention that for many circumstances, this extension is reasonably attained by the simple substitution of an image intensified camera. There are commercially available intensified cameras which should yield gains approaching 103 which is adequate to transfer the camera's operation from daytime levels of 103 lum/m2 down to quarter moon levels of approximately 10-2 lum/m2. Thus sensor sensitivity enhancements, even without exercising the camera's inherent injection inhibit mode, seem adequate to address all moonlight illumination levels down to at least quarter moon. For lower light levels, i.e. starlight only, additional technique development will be required, and may entail more extensive hardware modification. Based upon the nighttime sky and terrain radiance measurements (Duntley, 1970), there is good evidence that the spectral nature of the nighttime, moonlit sky is similar to that of the daytime sky down to about quarter- moon illumination levels. As this is readily verifiable using the radiometrically and geometrically calibrated whole sky imager itself with only minor optical modification, then the extension of our blue/red ratio technique into much of the nighttime regime is straightforward. Computational validations using the Hering FASCAT code, (Hering & Johnson, 1984), in a To the extent that the directional nature of the path radiance, induced by either direct solar irradiance or lunar irradiance plays its proportionate part along any selected horizontal path of sight, then the basic tenents of image transmission through the atmosphere (Duntley 1948) should hold true. It is our basic contention that this condition exists for all moonlit night skies, and that, therefore, contrast transmittance measurements, i.e. visibility determinations, derived from an intensified, passive imaging system are reasonable and practical to achieve even at the implied low light levels. Illuminance scaling to the nighttime flux levels should be well behaved under conditions where target selections and horizon sky backgrounds can be located without contamination by uncontrolled light sources. When near new moon illuminations drop below the levels required for contrast discernment by the camera ********** ** ** ******** *** ************* manner similar to that used for the daytime case, but substituting lunar irradiances rather than solar as input will provide additional insights particularly with respect to the influence of subtle shifts in the night sky spectral radiance distributions. 2.4 Analysis Considerations ***** ****** *** The analysis of the WSI cloud component's hardware performance can be addressed in conjunction with items a through f above by merely adding several WSI specific items in addition to the HSI specific items b&c. It should be clear that items a, d, e & f relate to mutually shared features of the HSI/WSI system and impact both transducer sub-assemblies equally. WSI hardware performance can be assessed by adding the following considerations: g) Sub-system's spectral selection reliability, i.e. is the filter changer sequencing properly? There are several inter-related aspects of HSI/WSI system performance that should be specifically considered in assessing the systems overall suitability. For convenience they might reasonably be separated initially into two categories. The first, those features associated primarily with the hardware performance, and second those features associated primarily with the decision algorithm performances. Obviously these two categories are not mutually exclusive, however they provide an initial framework upon which a comprehensive test procedure can be built. 2.4.1 Hardware Performance Analysis Sub-system's stray light control, i.e. is the solar occultor tracking? 1) Sub-system's flux control selections, i.e. are whole sky on-scale radiances optimized? The inclusion of items g, h and i in the overall system performance analysis should enable the definition and specification of a Hardware Figure of Merit that would assure the collection of an adequately defined imagery oriented data base to enable an equivalent and reliable analysis of the associated decision algorithm performances. b) The analysis of the HSI hardware performance should be based upon reasonable tests related to: a) System's reliability in responding to operator interrogation, i.e. does it come up when turned "ON" and initialized? System's directional reproducibility in locating target ROI's, i.e. do targets consistently fall within designated ROI's? c) System's radiometric stability, i.e. do video and auto iris circuits hold horizon radiance near predetermined level? d) System's consistency in grabbing & digitizing scene imagery, i.e. does the frame grabber capture at each scene? System's reliability in presenting updated computer products, i.e. does the decision software do as its told? f) System's environmental control sufficiencies, i.e. do hot/cold/wet/snow conditions cause mechan- ical, electrical or optical failures? An objective evaluation of the HSI's ability to satisfactorily meet the performance criteria listed in items a through f should allow a figure of merit, or statistically acceptable down-time specification to be defined. An analysis of past performance in these areas might readily yield an initial estimate, but an extended daily operational sequence, under either automatic or interactive control would yield a more suitable data archive. Since the existing composite HSI/WSI systems are daytime only devices which are built around realtime video data collection, it should be noted that the currently emerging night time capability, which is predicated upon the joint application of camera injection inhibit techniques and micro-channel plate intensifier technology, will require additional specialized testing. At the present time, Dec '89, the intensified cameras are in breadboard configuration. They are not ready for field operation, although many of the mechanical and electrical modifications necessary to retrofit the existing housings are either complete or in fabrication. Their operational concept was discussed briefly in Sections 2.3.3 and 2.3.4. Initial Laboratory testing of the intensified camera hardware sub-assemblies will involve an extension of items c, d and i in the discussion above. 2.4.2 Algorithm Performance Analysis The analyses of the HSI decision algorithm processes could be considered in several contexts since machine operations are always to some degree or another jointly influenced by operating system software characteristics and the philosophy underlying the intended logical sequences. For the application addressed herein it will be assumed that MS DOS is our 10 i.e. what relationships can be made to relate HSI, transmissometer, FSM, and human observer estimates? domain and that within the scope of the operations called for, i.e. described in Technical Note 213, the computer housekeeping and control software is adequate, and need not be tested further. Changes in software to improve user convenience, display options, etc., are not deemed necessarily as impacting the validity of the systems output products, and thus should be treated as off-line de-bugging exercises. This assumption is not intended to preclude the necessity for testing those software elements required to interface existing data formats with those subsequently specified by the sponsor. The algorithm performances then to be addressed within this test plan, are those whose decision elements transform digital imagery into numerical estimates of path of sight contrast transmittance, and thence sector and prevailing visibilities. It is presumed that the relationships discussed in section 2.3 and their supporting documentation listed in section 5 are an acceptable basis for defining visibility, and that, therefore, the test plan should directly address efficacy with which the HSI/WSI system provides adequate inputs to Eq. (2.1), and how the subsequent solution of Eq. (2.1) using these inputs can be externally validated. The analysis of the HSI decision algorithm performance should be based upon reasonable tests related to: The analysis of the HSI decision algorithms performance at night should at this stage of system development be based upon the premises defined in section 2.3.3, and thus ask two questions. First, can the intensified system extract usefully accurate apparent contrast values from the designated imagery, and thus under moonlight conditions continue to provide adequate data input to Eq. (2.1). And second can the intensified system find and identify within its surrounding scene an adequate number of suitably stable light sources to enable reasonable transition to radiance transmissometer mode? The answer to question number two will be highly site specific and the best night site is unlikely to coincide with the best day site. Airport sites where runway lights are within the cameras field of view will probably provide the optimum compromise. i) a) The operators ability to select enough targets, i.e. does the test site provide targets well distributed in range and azimuth? (ref. 2.3.1a) b) The operators ability to select appropriate targets, i.e. are available targets dark enough? (ref. 2.3.1b) c) The impact of insufficiencies in items a) & b), i.e. what happens when site provides only a few non-dark targets and Inherent Contrast estimates are poor? d) The appropriate detection threshold selection, i.e. for comparison with other machines or human observers. (ref. 2.3.1c) e) The impact of non-clear horizons, i.e. what happens when the weather gets lousy? (ref. 2.3.1d & e) The systems precision in extracting apparent contrast values, i.e. what are the least count and standard deviation characteristics within the imagery? The comparison of HSI visibility determinations with alternative techniques, Thus analysis of the HSI night algorithm performances should be based upon reasonable tests related to items "a" through "g" above, plus several additional items: h) At what flux level (illumination level) does the intensified system fail to yield adequate values of apparent contrast for selected target/horizon pairs, i.e. do normally occurring artificial light fields overpower natural moonlit scenes to the extent of precluding classical contrast transmittance measurements? What are the detection distances achievable by the HSI at night using 1) lights of opportunity as targets & 2) using lights known to be closely similar in intensity? Is the existing daytime field of view (FOV) adequate for nighttime sub-pixel detection of light sources? Is the detection of moderate intensity lights (25cd) at fixed distances an adequate calibration standard for night visibility, i.e. can these measurements be consistently correlated with the measurements in item "i"? An analysis of the WSI cloud cover decision algorithm for daytime skies was discussed in GL-TR- 89-0061, and previously introduced in Technical Note No. 210 in a separate, stand alone application. The technique is considered mature for total opaque cloud cover determinations, and is being refined to improve g) The 11 ** wym www.xewn i o w WwwWw . wwwwwwwwwwwwwwwwwwwwwwww wwww.SKARAR E AW ARAW/1/2kw.xkluwwwwwxxxwwwwwwwwwwwwwwwwwWwWWWWWWWWWWWWWWWWWWWWW WWWWWWWWWWWWWWWWWWWWWWWwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwww. . . .... ......... ..... its ability to discriminate cloud from aerosol forward scatter near the solar disc, and thin cirrus. For HSI/WSI applications the WSI cloud cover decision algorithm performance should be based upon reasonable tests related to: a) Comparison with total cloud cover determinations made by human observers, i.e. Can the WSI match or exceed the performance of human estimates? b) Estimated impact of restricted Field of View, i.e. does ignoring near horizon clouds severely bias total cloud cover estimates? c) Estimated impact of misidentifying the opaque to thin transition, and the thin to clear transition, i.e. does an uncertainty in the definition of "thin" cloud severely bias total cloud cover estimates? The analysis of the WSI cloud cover decision algorithm performance at night should be based upon the premises defined in section 2.3.4 and should, therefore, be based on tests related to the following additional items: d) At what illuminance levels does the blue/red ratio cloud detection process fail due to changing night sky spectral distribution, i.e. what happens when there's no moon? e) Under starlight only conditions is a radiance threshold algorithm adequate for cloud discrimination? Under starlight only conditions are subpixel starfield detections adequate for cloud discrimina- tion without additional FOV magnification, i.e. what is the optimum FOV for starfield detection as a cloud discriminator? Whi s !" """" . . ... . .... ...... .. . ..........S A ..... w w wwwwwww wwwwwvwn. **V.. ....... .. ...RASA Duc W ..Ve ""WW W WW W WAAI W ww 3.0 SPECIFIC TEST SEQUENCES The preceding two sections of this note have described the basic physical configurations of the composite HSI/WSI visibility cloud cover sub-systems, discussed the underlying premises upon which their automatic discriminations are predicated, and outlined specific performance criteria upon which the systems overall viability might be assessed. This section will outline a specific test sequence proposed for providing the data suitable for arriving at this assessment. ww ................. ... .. .. . . '' ' ' '' '' ' ' I il ".... ...!!.. .. ......w!..... . ......... ... ... FIELD TEST PLAN PROCEDURAL OUTLINE 3.1 Test Site Selection Requirements HSI: 1. Unobstructed horizontal paths of sight to the local horizon. (Optimum 360°) Heavy duty vibration free, geometrically stable platform. 3. Physically identifiable targets in all four quadrants (multiple targets per scene). 4. Scene & quadrant targets at distributed ranges (< 3 miles, 3-10 miles, 10-20 miles, > 20 miles). 5. All targets darker than horizon (preferably less than 10% reflectance). 6. Suitable network of support equipment (multiple sited scatter meters or transmis- someters). 7. Availability of experienced weather ob- server. WSI: 1. Unobstructed view of local sky dome (zenith angles > 80° not required). 2. Stable, level mounting platform (same as item 2 above). 3.2 HSI/WSI Hardware Performance Testing Perform the following operational checks on an hourly and/or daily basis until an adequate sample has been created to comply with G. L. specified reliability standards. HSI: 1. Power up sequence a. Computer system boots b. ACP activates & manual mode operates 2. Target/ROI display holds alignment 3. Horizon byte value held within range by Auto-Iris Frame Grabber (RGB line) digitizes each scene 5. Computer updates screen at each stop 6. Scan & Video normal under all weather events 7. Window contamination acceptable within ROI's WSI 1. Optical filter sequence runs normally 2. Solar occultor holds position 3. Flux control filter/Aperture sequence normal 4. Computer updates screen at each decision interval Computer: 1. Follows boot sequences per TN 213. 2. Implements program VISTEX to enable tape archiving. 3. Executes interactive interventions. 4. Executes automatic sequences as directed. 3.3 HSI/WSI Algorithm Performance Testing 1. Conduct standard AUTO-VIS runs daily in full automatic mode during each of several 2-3 day periods of meteorologically dissimilar conditions, i.e. clear, hazy, scattered cloud, full over- cast, rainy, snowing. Archive all data to tape! Computer: 1. Clean office environment within 100 ft or less from HSI/WSI external assembly. 2. Clean, relatively un-interrupted electrical power, 110VAC, 60hz. 13 .. d'. . . wi. . sv . . . . ... ... . .. X . X .wW w w . ww w ...!! w ..... .... . .. ..... .. ...... . ... V 6. En 2. Conduct simultaneous data collection se- quences using support equipment network as available. Archive. 3. Conduct simultaneous surface weather ob- servations as specified by NWS from MF1- 106(4-86), or equivalent, and retain written record. 4. Subsequent to each 2-3 day data interval, re-set the HSI systems initializing best estimate specifications of inherent contrast, Co, to new values, Co £ 5%, and run one additional data day at each of the two new values. 5. Perform & Archive such special interactive data collection sequences as deemed appro- priate by on-site test director. Extract from the archive accumulated in items 1-4 above, the data required to per- form the following evaluations. HSI: a. Sort by weather type 1. clear, VIS > 10 miles 2. hazy, VIS 3-10 miles 3. poor, VIS < 3 miles b. For each clear hazy & poor set 1. Time plot HSI vs FSM, TRANS, OBS 2. Scatter plot" C. For each A Co within item b sets 1. Time and scatter plots as in item b d. Sort A Co data sets for AOS, A Season 1. Plot impact of solar angle 2. Plot impact of seasonal variation in target refl. e. For optimized "clear" sets 1. Extract target byte values & std. deviations 2. Extract horizon byte values & std deviations 3. Calc impact on derived values of VIS 4. Compare against estimated errors in Co 5. Specify net figure of merit a. Extract 10 min cloud decisions from archive b. Extract hourly bogey red images C. Visually assess Red/Decision match Generate histogram comparison e. Extract Form 10b data f. Time plot WSI vs Form 10 3.4 HSI/WSI Nighttime Modification Since the ITT intensified chip that converts both the HSI & WSI sub-systems into their nighttime configura- tions is still in breadboard testing for noise rejection and RFI interferences, it is premature to be overly specific in outlining final test procedures. However, based upon the premises of sections 2.3.3 and 2.3.4, the night time test data base would be expected to include imagery equivalent to that outlined in section 3.3 above, plus several specific nighttime supplements. For HSI night imagery: 1. Add scenes including runway lights 2. Add scenes including distant street lights 3. Add scenes including distant lights of op- portunity For WSI night imagery: 1. Add scenes including unobscured starfield a. clear high desert site b. hazy coastal site Suey 14 .. ..... .. . . . 4.0 SYSTEM OPERATIONAL CONSIDERATIONS . . . .. ... . . .. ..... . ... . . . . . . . .. .. . . . . .... . . . . . .. maintains camera housing temperatures within reason- able bounds i.e. < 35°C, during the summer months, and its in-line liquid heater provides protection against sub-zero temperatures during winter months. The heater and chiller sub-systems are reasonably passive, thermostatically controlled devices which are relatively trouble free. Both HSI & WSI housing temperatures are displayed on the composite system ACP and should be monitored regularly by on-site personnel. Housing temperatures are also displayed on each digitized image that is viewed or stored to tape. User's should be aware that the CIDTEC cameras used in the HSI/WSI system are rated at 0° to 50°C and as usual, it is the overtemp condition that is most detrimental to it's health. Periodic inspections to maintain the coolant levels are recommended. 4.3 Manual Mode/Automatic Mode Procedures .... .**.**..* . . ... .... . . . . . . . . . . . . . .. . ..... . .! ! ! !! ! ! !!! !! ! !! l al' WWII' I ' . . ... . . . There are several operating procedures and house- keeping considerations that should be commented upon if the overall test plan outlined in section 3.0 is to be properly implemented. While these items do not effect the underlying technical principles upon which the system's procedural validity depends, they do impact the efficiency and reproducibility with which the test sequences can be conducted and later intercompared. In this context they are therefore primarily addressed to the field experimentalists concerned with the physical set- up and execution of the on-site test sequences. 4.1 Set-up Requirements for Physical Stability It is important to recall that the HSI operating concept is built around the consistent re-acquisition of distant targets within predefined regions of interest. The larger the number of suitable, range distributed targets that can be specified, the better the system will perform. Thus sooner or later, the user will be driven by site specific limitations to choose angularly small targets in order to optimize his target suite in each sector. Since the HSI field of view (FOV) is 5.4°, the nominal per pixel FOV is approximately 0.01 degrees, or about one foot at a distance of one mile. It is to be expected that with resolutions of this sort, the user often will be inclined to push the target selection process to its limits. It is at this point that mechanical instability in the HSI mounting structure can defeat the optical and radiometric capabilities of the device. To accurately reproduce target and ROI coincidences within 1 pixel, it is essential that the WSI housing not move under the influence of wind blast, wetting and drying of wooden support structures, or other flexible mounting fixtures. The metal trapezoidal support frame illustrated in Fig. 1-7 must be mounted on an inflexible, rigid base. The degree to which this is accomplished will determine ultimately the minimum size target that can be used effectively in the system's automatic sequences. 4.2 Environmental Protection Devices The composite HSI/WSI system is designed to acquire, digitize, manipulate and archive image oriented data in a fully automatic, relatively unattended manner. These procedures are normally exercised under software control by code resident within the 80286 based microcomputer sub-system. Operation in full automatic mode requires only that the system be initialized with program SETUPVIS as outlined in Technical Note 213, and then passed off to program VISTEX. ..... .... ... .. ... . . . . . . .. .......... ...... As a diagnostic and trouble shooting tool, the system's accessory control panel (ACP), see Fig. 1-6, enables operator intervention into most peripheral sub- system controls. Thus, in the event of anomalous system performance at either hardware or algorithm levels, the operator can switch the ACP from "REMOTE", i.e. Computer control, to "LOCAL", i.e. manual control, and exercise those sub-systems suspected as the source of the anomaly, i.e. HŞI rotary table, WSI filter changer, etc. In a similar manner, the Manual Mode can be used to establish fixed non- standard control settings for conducting special purpose measurement or test and evaluation sequences. Caution should be exercised to insure returning the ACP switches to "REMOTE" prior to reinitiating automatic operations. 4.4 Data Archival and Recovery .. . .......... The HSI/WSI composite system is designed as an all weather system, but has not been packaged to totally defeat every conceivable weather condition. It is designed to remain operationally undamaged under most mid latitude weather scenarios. Its waterproofing O-ring type assembly and internal desiccants protect it against rain and moisture. Its liquid chiller sub-system The composite HSI/WSI system has been designed to provide several archival options. Small amounts of imagery can be held on the Seagate 45 Megabyte hard disc via program SAVIMG, which allows relatively 15 . ... web quick interactive retrieval and manipulation of operator selected data. Larger amounts, representing several days worth of continuous data collection are normally diverted to the EXABYTE 2.2 Gigabyte streaming tape sub-system under program VISTEX control. Retrieval from this tape archive is normally done off-line for more extensive quality control and analysis procedures. Several specialized software routines are required for executing these data retrieval options. Normally program VIEWVIS is used to manipulate the EXABYTE archive, but is run off line and not necessarily exercised at the field sites. Procedural instructions and support hardware requirements necessary for efficient EXABYTE data retrieval and manipulation will be provided under separate cover. performance versus that of the human is probably only measurable in a statistical sense, since humans are so inconsistent and relatively slow to train. Comparison against other more simplistic hardware systems is easy but hardly definitive given the inherent limitations of scatter meters and transmissometers in determining the directional impact of path radiances. An extensive set of image data illustrating algorithm sensitivity to input parameters and measurement precisions is our best recommendation for establishing the systems overall figure of merit. .. .. .. .. .. .. ... .......... ....... 5.0 SUMMARY COMMENTS . ... ... .. . . The proposed test plan and supporting technical discussions presented in this document have been provided as a guide to the orderly and realistic evaluation of the performance characteristics of the composite HSI/WSI imaging system. The background material presented in sections 1 and 2 has been selected in an attempt to provide a single, complete source for not only the proposed stepwise test procedure, but also for the material deemed useful in understanding the technical basis for the specific procedural selections. One concluding comment with respect to the determination of prevailing visibility seems appropriate. The majority of the discussion in this note has been directed toward defining the efficacy of determining the contrast transmittance, i.e. visibility along a single path of sight, or along several closely grouped paths contained within a single scene or visual fixation. The techniques discussed herein are deemed appropriate and specifically applicable in this context. However, the concept of "prevailing" visibility implies a relatively complex sorting and combining of these discrete determinations into a subtly defined composite reflecting a somewhat subjective spatial assessment. As it turns out, a conceptually accurate determination of prevailing visibility requires a balanced distribution of appropriate targets in each azimuthal zone, a circumstance not normally achieved in the real world. A human observer copes with this by ignoring the rules defining "visibility" and uses "any" object he/she can see and identify in each scene to estimate the detection range. In doing so, the observer has violated the underlying tenets implied in the Koschmieder relationship, which the video based system has been purposely required to obey. This fundamental contradiction is being addressed currently by the HSI system via a relatively simplistic prevailing visibility algorithm. It does not fully address the problem created by anomalous target distributions, but provides a first step. Continuing research in this area is underway and will be discussed further under separate cover. There is good evidence to support the contention that many of the features included in the test plan have been adequately addressed during the past year of field operations. This is certainly true with respect to the hardware reliability of the WSI sub-system, which has been continuously operational at seven CONUS sites for up to two years with only minimal host support. The field operation of the HSI sub-system has been less intensive, although two units have been intermittently operational at the GL & MPL sites for about a year. Hardware reliabilities are addressed in the proposed plan however to insure completeness and provide a mechanism for updates or additional items where deemed appropriate. The validation of the decision algorithms, and determining their sensitivities to the uncertainties inherent in their required input parameters, is a task which will be driven by the sponsors specification of "ground truth", and requirements of the end item's application. The daytime HSI/WSI system has been developed with the intent of duplicating the performance characteristics of the human observer while improving upon his consistency and precision. The systems 6.0 REFERENCES & BIBLIOGRAPHY Douglas, C. A. and Bookker, R. L., (1977). "Visual Range: Concepts, Instrumental Determination, and Aviation Applications", National Bureau of Standards, NBS Monograph 159, FAA-RD-77-8, ADA 041098 Duntley, S. Q., (1948). "The Visibility of Distant Objects", J. Opt. Soc. Am. 38, No. 3, 237-249. 16 . . .... ......... . Duntley, S. Q., (1948). "Reduction of Contrast by the Atmosphere", J. Opt. Soc. Am. 38, 179-191. . .. ments with a Solid State Imaging System", University of California, San Diego, Scripps Institution of Oceanography, Marine Physical Laboratory, SIO Ref. 89-7, GL-TR-89-0061. .. ...! !!! !!!! ?? ? h r ipri , Mini, M w i l l . Duntley, S. Q., A. R. Boileau, and R. W. Preisendorfer, (1957). "Image Transmission by the Troposphere I", , J. Opt. Soc. Am. 47, 499-506. Gordon, J. I., (1979). "Daytime Visibility, A Conceptual Review", University of California, San Diego, Scripps Institution of Oceanography, Visibility Laboratory, SIO Ref. 80-1, AFGL-TR-79- 0257, NTIS No. ADA 085 451. Karr, M. E. and Johnson, R. W., (1989). "Horizon Scanning Imager, Preliminary Operations Manual", University of California, San Diego, Marine Physical Laboratory, Optical Systems Group, Technical Note No. 213. '" ' ' , "' "** .. .. .!! !! .. ..!! INI M A.. ...... .. ........ . .... . . Hering, W. S., Muench, H. S. and Brown, H. A., (1971). "Field Test of a Forward Scatter Meter", AFCRL-71-0315, Environmental Research Paper No. 356. ....... .. Pandiwsisiw...w.man ........... ... .. Hering, W. S., Muench, H. S. and Brown, H. A., (1972). "Mesoscale Forecasting Experiments", BAMS 53, No. 12, 1180-1183. .. .... . ............. ... .......... ***/.../... W W ww .S Hering, W. S., (1981). "An Operational Technique for Estimating Visible Spectrum Contrast Trans- mittance", University of California, San Diego, Scripps Institution of Oceanography, Visibility Laboratory, SIO Ref. 82-1, AFGL-TR-81-0198, ADA 111823. n. Awwwwwwwwwww..!!!... atw. WIVAAW . ./... w? s w ., Iliw; AWNI WIMWI AMA wwwvinilawiwibWwwww w /.. .../A. . iwww.wiwivis / wwwwwviw i liw W Hering, W. S. and Johnson, R. W., (1984). "The FASCAT Model Performance Under Fractional Cloud Conditions and Related Studies", University of California, San Diego, Scripps Institution of Oceanography, Visibility Laboratory, SIO Ref. 85- 7, AFGL-TR-84-0168. Hering, W. S. and Johnson, R. W., (1988), "The Determination of Atmospheric Visibility with a Digital, Image Acquisition System", University of California, San Diego, Marine Physical Laboratory, Optical Systems Group, Technical Note No. 209. Johnson, R. W., (1981). "Daytime Visibility and Nephelometer Measurements Related to its Deter- mination", Atmospheric Environment 15, No. 10/11, pp. 1835-1845. ANITA www.hwprawiwwwwwALAF W vwww.www .hivy:n ...... ......... ... .. ... . . . All I W all Aviv.PA .... .. ..... .. **** W AWA ... Johnson, R. W. and Hering, W. S., (1988). "Automated Visibility and Cloud Cover Measure- ments With a Solid State Imaging System, The As- built First Generation", University of California, San Diego, Marine Physical Laboratory, Optical Systems Group, Technical Note No. 207A. Johnson, R. W. and Hering, W. S., (1989). "Automated Visibility and Cloud Cover Measure- . . . . . .. ... ... ... ... .. . . ... . ....... .. 17 ???../V.... .. . ..* *222 ** *. *. ...... ... www. w w wwww