UNIVERSITY OF CALIFORNIA, SAN DIEGO 3 1822 04429 7216 JUNE 1986 ATMOSPHERIC VISIBILITY TECHNICAL NOTE NO. 202 AN AUTOMATED VISIBLE SPECTRUM IMAGING SYSTEM FOR ATMOSPHERIC MEASUREMENT & ANALYSIS Offsite (Annex-Jo rnals) QC 974.5 . T45 no. 202 R.W. Johnson W.S. Hering 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 written consent of the Visibility Laboratory and Air Force Geophysics Laboratory. Contract Monitor, Donald D. Grantham Atmospheric Sciences Division TERSIT TRE 2NIA . TER Prepared for Air Force Geophysics Laboratory, Air Force Systems Command United States Air Force, Hanscom AFB, Massachusetts 01731 (1868 SCRIPPS INSTITUTION OF OCEANOGRAPHY VISIBILITY LABORATORY La Jolla, California 92093 UNIVERSITY OF CALIFORNIA, SAN DIEGO 11 LIT III D 3 1822 04429 7216 TECHNICAL NOTE NO. 202 AN AUTOMATED VISIBLE SPECTRUM IMAGING SYSTEM FOR ATMOSPHERIC MEASUREMENT AND ANALYSIS This Technical Note contains the extended summary and vu-graph representations that were presented at the Fourth Annual Tri-Service Clouds Modeling Workshop which was held at the Air Force Geophysics Laboratory, Hanscom AFB, Massachusetts, during the first week of June, 1986. The technical data were prepared by W.S. Hering and J.E. Shields. The workshop presentation was made by R.W. Johnson. vury_6 _OF_C COPIES AN AUTOMATED VISIBLE SPECTRUM IMAGING SYSTEM FOR ATMOSPHERIC MEASUREMENT & ANALYSIS Richard W. Johnson and Wayne S. Hering Visibility Laboratory University of California, San Diego Scripps Institution of Oceanography La Jolla, Califomia 92093 ABSTRACT The Visibility Laboratory, in support of several DoD sponsored programas, is developing a visible spectrum, computer controlled imaging system for the automatic assessment of the optical state of the atmosphere. The basic hardware configurations associated with both staring and scanning optical modes are described as related to their cloud discrimination or visibility deter- mining tasks. Fundamental calibration of the multi-spectral video imagery is outlined and sample data are shown for illustration. Since the goal of this multi-faceted task is the fielding of a fully automatic, unsupervised system, the development of adequate interpretive algorithms is by far the most exciting and potentially far reaching technical aspect of the program. Preliminary applications of this emerging technology have been directed toward the detection of cloud-free paths of sight and their related statistical characteristics. Examples of initial semi-automatic computer extractions from video imagery are shown and discussed, as are the underlying concepts upon which the decision algorithms are based. INTRODUCTION The proliferation of advanced military electro-optical (E/O) systems that are designed for visible spectrum operations within the troposphere has greatly increased the need for a compact automatic system designed to assess and predict the optical and meteorological properties of this operational medium. (Figs. 1,2,3). The Visibility Laboratory, under the sponsorship of the Air Force Geophysics Laboratory, is developing a family of small, solid-state imaging systems to address this need. Several existing and developing versions of these devices and their proposed uses are described briefly in the following paragraphs and illustrations. (Figs. 4,5,6). In its most basic form, each of these devices consists of a computer controlled solid-state video system that provides calibrated multi-spectral imagery suitable for the automatic extraction of local image transmission and cloud cover information. Imbedded within the control computer are prototype and proprietary extraction algorithms necessary to provide these numerical products. In addition to radiometrically calibrated imagery, advanced algorithm development is currently underway to provide near real-time products of the data acquisition, processing, and display system in the form of continuously updated digital presentations of selected operational quantities. The quantities most desired for describing the optical state-of-the-atmosphere in which this sensor system operates are, of course, task dependent, but current algorithm development is slanted toward cloud detection, cloud-free arc determinations, sector visibilities and total cloud cover. The multi-spectral imagery is equally applicable in more generalized search, detect and identify scenarios. INSTRUMENTATION The system block diagram shown in Fig. 4 is intended to illustrate configurational relationships common to all systems described herein, and is not necessarily an as-built drawing. For example, the "select" function linking the Z-100 computer and the Time/Date Generator is conceptual only. Both the video and digital data are routed through the system simultaneously and independently. The parallelism is an operator convenience, not a necessity. Most of the operational characteristics illustrated in Fig. 4 are included in each of the currently evolving hardware configurations. · The schematic drawing of the multi-spectral imager shown in Fig. 5 illustrates two important features of the camera system that contribute to its general utility. . First, the Filter Changer Assembly is shown to contain two independently controlled filter locations on the optical axis. In actuality, this assembly contains two filter wheels, each holding four separate filters in addition to the five lens optical relay. Each wheel can position any one of its four filters into the optical path under either manual or computer control. In its present configuration, the forward wheel contains four glass absorption neutral density filters, while the rear wheel contains four spectrally different interference filters. Thus, both spectral band and flux level can be controlled either manually or by pre-determined computer program Second, the multi-lens Turret Assembly (shown pictorially, not as built), provides an efficient method of modifying the overall optical path to meet task specific requirements. In the prototype systems currently in use, the two lens assemblies require manual substitution into a fixed adapter. The remotely controlled Turret Assembly is still under development, as are the remote control of the Iris and Occultor sub-assemblies. However, it is important to note that the system lends itself well to task specific changes in Turret design. The composite, as-built system shown in Fig. 6 is the currently operating Z-100 based system. Only the digital option previously illustrated is contained in the instrument racks. The camera, filter changer, and fisheye lens assembly, are supported on the temporary camera tripod. This successfully deployed operational system is designed primarily for image acquisition and archival only. The computer does not contain the programmed algorithms necessary for sophisticated image manipulation and analysis, although relatively simple data extractions and numerical computations are readily available. The system can reliably acquire an image every two seconds and store it on disc prior to their batch mode downloading to magnetic tape. The first generation systems shown in Figs. 5 and 6 both use the GE2505A2 solid-state CID video camera as the primary detector. This camera outputs a standard R$170 composite video signal which is grabbed and digitized by a Poynting 505 frame grabber. The resultant 8-bit 256x256 data array is subsequently operated upon to yield an image that is radiometrically calibrated in absolute radiance units traceable to the National Bureau of Standards. Geometric calibrations are performed on the array such that with the 174° Fisheye adapter lens in place the system yields an angular resolution of approximately 0.7 deg/pixel, and with the 30° small FOV lens in place yields approximately 0.1 deg/pixel resolution. SYSTEM CALIBRATION 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 a facility similar to that shown in Fig. The step-by-step sequences required to accomplish the calibration procedures outlined in Fig. 7, and create the calibration data summarized in Fig. 8, are discussed in detail within several separate notes and will not be addressed herein. The key feature underlying all of the calibration sequences is, of course, the stability and reproducibility of the system radiometric linearity. An example of the Linearity Calibration conducted on each system is illustrated in Fig. 10. These cameras are generally linear in their response to changing flux levels over a span of about 1.1 decades, as illustrated in Fig. 10, however, second generation imagers are becoming available with twice this dynamic range. The geometric calibration of wide field-of-view systems is simplified by the Visibility Laboratory facility illustrated in Fig. 11. This sixteen foot diameter dome contains 193 well-defined self-illuminating sectors. By placing a fisheye camera at its center, an image may be acquired that will clearly characterize the systems field-of- view. A sample image obtained in this facility by the video camera described above is shown in Fig. 12. Note that since this image is from a standard R$170 video camera, the characteristic 4x3 image aspect is transferred to the digital representation recreated in Fig. 12. A computerized correction to recover a more attractive circular image is straightforward but not normally applied. SYSTEM PERFORMANCE Substantial quantities of all-sky imagery have been collected that clearly indicate the efficacy of cloud detection and characterization using the system shown in Fig. 6. Examples of these data and their application to determining selected cloud field statistics are illustrated in Figs. 13-19, which have been abstracted from an in-house Technical Note titled "Imagery Assessment for the Determination of Cloud Free Intervals", May 1986. Typical cloud field imagery illustrated in Fig. 13 can be readily operated upon to yield the lineal radiance distributions shown in Figs. 14 and 15. In Fig. 14 data acquired by several competing hardware configurations, two electro-optical and one photographic, are compared to illustrate the comparability between the two higher resolution systems. The point to be made is that the geometric resolution of the 256x256 electro-optical camera is quite adequate, and for these applications it does not suffer in comparison with the slower, more tedious processing requirements of the photographic system. A primitive cloud/no-cloud decision methodology can be illustrated by comparing line grab data from two images obtained at the same place and local time, but on separate days under different sky conditions. Such a set of circumstances are illustrated in Fig. 15 which compares line grabs from the images in Fig. 13a and 13b. It is implied in Fig. 15 that a simple radiance thresholding algorithm will satisfactorily detect those locations along the prescribed sky arc where clouds occur, and where they do not. In many cases, this implication is true, however this simplistic procedure often needs help. Two examples of machine discrimination are shown in Fig. 16 where the original monochromatic image taken at 645nm is compared with a simple radiance threshold recreation and an edge gradient supplement. It is apparent that these straightforward techniques reproduce the original cloud field distributions with adequate fidelity for most statistically driven analyses. Application of the automatic processing algorithm illustrated in Fig. 17 rapidly produces output products such as those shown in Figs. 18 and 19. These sample plots represent the comparative results from an analysis of 233 hourly images conducted over a 23 day interval ending 20 September 1984. The statistics were developed via the algorithm of Fig. 17, manually by visually identifying cloud/no-cloud arcs on the display monitor and via model applications following the procedures of Allen and Malick. The results of this preliminary test are encouraging, and we are proceeding with the development of automatic techniques. There are, of course, shortcomings related to the use of monochromatic imagery. They are in general well known and are summarized in Fig. 20. The most straightforward procedure to circumvent the problems associated with monochromatic image interpretation is to use multiple wavelengths. For cloud oriented applications we have found that paired imagery using blue and red passbands is an effective procedure. Figure 21 illustrates a scene recorded in four visible spectrum bands at approximately two second intervals upon which our current blue/red detection technique has been based. 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 graphically in Figs. 22 and 23. In these figures, the blue/red sky radiance ratio has been plotted versus zenith angle along an east-west horizon to horizon sky arc for each of several surface visibilities. Equivalent blue/red ratios for clouds fall near the 1.0 line. Current data base searches are directed toward defining the real world variance of the cloud ratio from its characteristic value near unity. The eastern sky near sunrise, Fig. 23, continues for example to be a difficult region, however, the blue/red radiance ratio will surely continue as the basis for a reliable, consistent cloud detection methodology. Two final illustrations comparing the derived blue/red sky radiance ratio images with the original mono- chromatic red (649nm) images are shown in Figs. 24 and 25. Figure 24 illustrates the imagery in the context of an all-sky fisheye field-of-view, while Fig. 25 illustrates the same technique applied to a higher resolution 30° field-of- view. Not only are simple cloud/no-cloud discriminations clearly defined, but a variety of more sophisticated cloud edge characteristics related to sub-saturation optical depth determinations are readily discernible. From our experience to date, it is clear that a small, relatively automatic system for the determination of cloud cover and related statistical characteristics is now available to the experimental and modelling communities. Figure 26 illustrates our conception of an automatic system designed to accumulate the imagery required to address not only the specification of local cloud cover, but also the effects of directional contrast transmittance, and related optical phenomena. The application of this prototype system to the task of multi-purpose, local area measurements will begin shortly and should be the harbinger of a new standard of consistent and reliable weather related observations. VISIBILITY Figure 1 VISIBILITY VISIBILITY VISIBILITY VISIBILITY VISIBILITY VISIBILITY VISIBILITY VISIBILITY LABORATORY Scripps Institution of Oceanography University of California, San Diego SITUATIONS, REQUIRING ATMOSPHERIC_STATE SPECIFICATIONS 1. AIRBORNE AND SURFACE TACTICAL PLANNING 2 HOSTILE ENVIRONMENT MONITORS 3 UNMANNED AIRFIELD REPORTS 4. METEOROLOGICAL REPORTING AND FORECASTING S. MODEL EXTENSIONS TO NON-UNIFORM CONDITIONS Figure 2 COMPONENTS OF ATMOSPHERIC_STATE SPECIFICATION 1. VISIBILITY 2 CLOUD COVER I CLOUD BASE HEIGHT 4. RAIN - SNOW . OBSCURANTS E/O CAMERA III - DISCUSSION OUTLINE Figure 3 SYSTEM SCHEMATIC & BASIS CHARACTERISTICS RADIOMETRIC CALIBRATIONS SINGLE-COLOR CLOUD DETECTION PROCEDURES 4. IMPROVED TWO-COLOR ALGORITHM DEVELOPMENT 5. ADVANCED TECHNIQUES FOR EXPANDED APPLICATIONS PROTOTYPE SYSTEM AS-BUILT FOR AUG '84 DEPLOYMENT PANASONIC SOLID STAFF CCTV MONITOR [R.989 E/O CAMII IN 2505 CAMERA ASSY FOR VIG-22 TIME/DATE GENERATOR IRIS ASSY OXCULTOR ASSY SONY VOU) 3/4 VOR REMOTE PANEL IRIS EXCULTOR CONTROL CONTROL Figure 4 COMPOSITE SYSTEM PROPOSED FOR CFLOS ALGORITHM DEVELOPMENT VISITOR VTOFO (XCULTOR ASSY VIDEO VIDEO TIMEDATE GENERATOR E/O CAM III FILTER TN 2505 CHANGER CAMERA LSSY. -TT IRIS ASSY SOR VIDEO REMOTE DIGITAL POYNTING SOS FRAME GRAB ITS 100 CONTROLLER CIPHER 9 TRACK DIGITAL RECORDER FILTER CONTROL 1 FRIS ASSY OCCULTOR ASSY DIGITAL INTERFACER (HUJ ZENITH 2.100 COMPL'TER - - - - COMPOSITE SYSTEM PROPOSED FOR OPTICAL STATE OF THE ATMOSPHERE ASSESSMENTS Fisheve Lens Assembly (174° FOY) www Figure 5 Image Location . CULU General Purpose Lens Assembly (30° FOV) → Filter Changer Assembly 2505 Camera Assembly Multi-Lens Turret Assembly Figure 6 E/O CAM III: SYSTEM CALIBRATION PROCEDURES 250S CAMERA RADIOMETRICALLY UMFORM TARGET CHARACTERISTIQ O SNTZ со СССР DLFIMCO TANGOT CHARACTERISTIC SEL RET SHOTO POYNTING SOS FRAME GRAD RAW DIGITIZED ARRAY UNIFORMITY MAP CORRECTION SONO SWT 31 CORRECTED DIGITAL ARRAY TRUE RADIANCE DISTRIBUTION Intro RADIANCE ARRAY OR SWT O O SNT URO 7 SOU ENT 10 Figure 7 - - - - - - - -- - - - - - POLARIZER TRANSMITTANCE CARACTERISTICS SEE RES SMT6 RADIOMETRIC LINEARITY | CHARACTERISTICS SEL TO SHTE RADIOMETRIC AGS SENSITIYITY CHARACTERISTICS sa KO SATS - - - - - - - INPUT FROM HARDWARE OUTPUT TO TAPE PROCEDURAL RELATIONSHIPS WITHIN CALIBRATION SEQUENCE - - EIO CAM III: CALIBRATION TAPE CONTENTS - 1. RADIOMETRIC LINEARITY (relative flux input vs byte value output) 2. ABSOLUTE RADIANCE RESPONSE (absolute flux input vs byle value output) a. Componar Small FOV in each speclied å b. Bicar Fisheye in each specified à Figure 8 3. RELATIVE RADIANCE SHIFTS (relative flux input vs byle value output) a. Componar Small FOV: Avs f-stop & à vs N.D. 6. Bicar Fisheye : Avs 1-stop & A vs N.D. 4. GEOMETRIC MAPPING a. Hemisphenc Dome: Fisheye 6. Keyhole Room: Fisheye Small FOV C. Linear Wall Scale: Small FOV 5. ARRAY UNIFORMITY DATA a. Extracted from items 2 & 4 ABSOLUTE CALIBRATION PROCEDURE QUALE STANOAAD OF AMOUNT NTENSITY W 28547 CALIBRATE TAAGET Ape 8,50 LAMP EO CAMERA ASSEMBLY AOUNO TARGET Figure 9 - CONTROL COMPUTEA 3 METEA OPTICAL BENCH & ACCESSORIES TAPE ARCHIVE 11 AEGULATED POWER SUPAY logy COS E/O CAM III: LIN CAL CAM 436-1854 GRABBER 84 06 02 30 SEP 85 RELATIVE FLUX Figure 10 50 100 - 150 200 250 250 300 BYTE VALUE OUTPUT Figure 11 i * ..... - GEOMETRIC CALIE • DOME Figure 12 a) Clear sky b) Cumulus clouds Figure 13 c) Alto cumulus d) Cimus clouds Fig. 4-1. Imagery from EO Camera il for a variety of sky condtione Images were digitized from video, Dipod through image processing. displayed into matru camera (a) EO Camera Il. (see Fig. 4-10) (b) EO Camera 1. (see Fig. 4-2a) 150r 256 SIGNAL SIGNAL 128 512 64 128 256 PICTURE ELEMENT (Pixel) NUMBER PICTURE ELEMENT (Pixel) NUMBER (c) Automax l. (see Fig. 4-2b) 100 Figure 14 SIGNAL 256 1 28 PICTURE ELEMENT (Poxal) NUMBER Fig. 4-3. Line plot of cumulus cloud image for EO Canera #. EO Camera I and Automax l. The data wera acquired 3 SEP 84 near 1401 hrs. Row 282 Irom EO Camera and corresponding rows of EO Camera I and Automax i were extracted. 150 - 13 SEP CAMERA SIGNAL 75 Figure 15 4 SEP 256 512 PICTURE ELEMENT (PIXEL) NUMBER Fig. 4-4. Measured signal (from radianca at 650nm) as measured by the EO system on 3 SEP and 4 SEP 1984 at 1400 LST along an east-west arc passing naar zenith. 3 SEP had broken cumulus, 4 SEP was clear. Fig. 1-5. IST OF RADIATICE THRESHO!. Ciriciclica.ii:.:: ORIGINAL IM som O 1O 1401 met twee ELS witá i la) Original image. 101 image 27.01. 1977:*:!P!5 2015 :7777 50 c0.07 C00:05wt? Fig. 4-6. GRADIENT INFORMATION IN CLOUD DETECTION TESTLJAGE. SEPS: 40: mu HDMI PRIMO FIT Clouri IN DIY with automoto . 1.7-03-CTd1: ܫ ܢ ܫ ܪ ܫ ܝ ܚ ܀ ܝ (a) Pseudo gradient derived from image. 01 Pseudo 6:00.er: coor.cocedi-C. Sem00sec On"?E"CCCCC: cocec mise Figure 16 _ START TESTCLO Tapes of FO images 512x512 Read mago, extract 1 row. Jssociato win tardato file CLOUD.X file of dont who he time, etc file CLDARC 1 row from each wage. with dent into ENO TESTCLD Aoad CLOARC, acoly cloud algonthm. cid-1. nocido STAAT CLOSTATS ilio CLOBIN apply computed e for each * binary live. O&1, clo resuns as a tunction of A Figure 17 Accumulate results for each row, compute treg. CFLOS vs file CLOSTAT Deloa longest cld troe arc. each imago, accumulate results of cum tred of ald froe arc length CFLOS vs o. cumtree. cle ire. arc length vs arc length ΕΝΟ CLOSTATS Fig. 5-1. Conceptualized chant of EO Camera dau processing program FREQUENCY CLO FREE LINE OF SIGHT 100 FREQUENCY (%) AUTOMATIC MANUAL --- MODEL Figure 18 -80 -60 -40 -20 0 20 40 60 80 OBSERVATION ZENITH ANGLE Fig. 5-3. Cloud frow line of sight, frequency of occurrence na zunan angre. CUM FRED CLO FREE AAC LENGTH CUMULATIVE FAED (%) Figure 19 AUTOMATIC --MANUAL 20 40 60 80 100 120 140 160 ARC LENGTH (DEGREES) Fig. 54. Cloud free arc longin, cumwalme frequency of occurrence vs ac longin. SINGLE WAVELENGTH DISCRIMINATORS: CAVEATS 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 2. EDGE GRADIENT TECHNIQUES a. Cloud Imbedded in Cloud Background Yields Redundant Boundaries b. Fails as Scene Approaches Uniform Overcast Figure 20 is. ':'::';1 '.. " 4SSNM W . MEASURED SKY BRIGHTNESS EO CAMERA SYSTEM FISHEYE VIEW 64SNM Figure 21 BLUE-RED SKY RADIANCE RATIO Visibility Equinox 10:00 AM Lat 30 N T RADIANCE RATIO West East + -80 -60 -40 -20 0 20 ZENITH ANGLE 40 60 80 Figure 22 BLUE-RED SKY RADIANCE RATIO Visibility Equinox 7:00 AM Lat 30 N RADIANCE RATIO West East -80 -60 -40 -20 0 20 ZENITH ANGLE 40 60 80 Figure 23 ST NA SA . .. BLUE RED BRIGHTNESS RATIO FISHEYE VIEW Figure 24 64 SNM ', I BLUE RED BRIGHTNESS RATIO NARROW (381EG FOR Figure 25 PN 5*SIM Automatic Observing System for Whole Sky and Horizon Imagery Occulter Fisheye Lens Shutter Assembly Telephoto Lens Target Scene ..: SX33XX Heated Observation Windows Kde 43 D SWM W . VOOR3 SEMERXX tikel 2Y- - Horizon * * . VA . * . X ** . . 1 Filter Assembly 50/50 Beam Splitter d, - d2 = 150 mm. ell. Telephoto di +d3 = 30 mm, ell. Fisheye L 114 i 11 - Local Horizon 11IH Camera Insulated Weather-Proof Housing . . To Control Panel Figure 26 - ، بس سر . م . می ء * . " . . م - ا . . . . تسس | -- سر " ا ا ما : - -- مي سے " کر دیا مسم . و میر .. . همسه. . ۱ (. م . . . . . .