978 PHOTOGRAMMETRIC ENGINEERING Asia and the Far East held at Bangkok, Thai­ raphy and Projection for Military Photo Inter­ land, October-November 1961. pretation." PHOTOGRAMMETRTC ENGINEERING, Wilson, R. c., 1962. "Surveys Applicable to Ex­ Vol. XXVII(3): 450-460. tensive Forest Areas in North America." Pro­ Woo, H. W., 1962. "Vela-Uniform-Task VII On­ ceedings of the Fifth World Forestry Congress held Site Inspection." Proceedings of the Symposium at Seattle, Washington, September 1960, Vol. 1: on Detection of Underground Objects, Materials, 257-263. and Properties held at Ft. Belvoir, Virginia, Winterberg, R. P., 1961. "Additive Color Photog- March 1962, p. 250. Human Factors in Image Interpretation* ROBERT SADACCA, U. S. Army Personnel Research Office, Washington 25, D. C. REPORT TO SUBCOMMITTEE III PHOTO INTERPRETATION COMMITTEE 1962-63 AMERICAN SOCIETY OF PHOTOGRAMMETRY 27 MARCH 1963 VEN a cursory survey of human factors E problems in image interpretation leads to the conclusion that the problems involved are manifold and extremely complex and that a broad, integrated program of research needs to be implemented. Today, interpreters are not only confronted with the problems of interpreting relatively large-scale, conven­ tional black-and-white photographs, bu t they are also required to interpret small-scale and degraded photographs as well as radar, and infrared imagery. In addition, recent techno­ logical developments have increased the capacity of reconnaissance systems to obtain and process imagery at a rapid rate. The imagery obtained through this increased capacity and variety of image sources is ulti­ mately placed before image interpreters who are asked to extract information to be used in important decisions. ROBERT SA DACCA Yet existing knowledge about the basic ceptual set, background information, stimu­ psychological factors operating when inter- Ius quality, and personality variables on the preters examine reconnaissance imagery is speed, accuracy, and completeness of inter­ severely limited. Relatively little is known preters' final decisions? These are human concerning the perceptual and cognitive factors questions; unfortunately, they are not processes that occur when interpreters search easily answered. imagery and make critical decisions concern- Image interpretation is an extremely com­ ing the presence or absence of significant plex process. Impinging upon an interpreter's objects. What clues or target signatures do performance are factors related to his back­ interpreters use? What inferential processes ground, training, aptitudes, interests, and do they employ? \;\That are the effects of per- personality, as well as his momentary state of * Presented at March 24-30, 1963 ASP-ACSM Convention, Hotel Shoreham, Washington, D. C. The opinions expressed in this paper are those of the author and do not necessarily reflect official De­ partment of Army policy. EDITOR'S NOTE: The Abstract for this paper is on page 482 of the 1963 YEARBOOK. HUMAN FACTORS IN IMAGE INTERPRETATION 979 motivation and fatigue. Factors related to must be clearly delineated and the efforts of amount, quality, and content of the input interested researchers must be pooled and imagery, time available for viewing the coordinated. Some concrete suggestions along imagery, the kinds of instruction and target this line will later be made. information the interpreter has received, and the techniques and equipment he uses in Research in the first area, identification of examining the imagery also play major roles basic human factors in image interpretation, in determining performance. Performance should have as its ultimate objective the itself can be measured in a variety of ways­ improvemen t of interpreters' productivi ty number of correct identifications, number of through the optimal use of their talents, skills, errors, per cent accuracy, per cent complete­ and abilities. Unfortunately, most of the ness, rate or speed of information extraction. psychological research into perceptual and And, of course, the task of the interpreter can cognitive processes that has been conducted vary. The task may consist of detection, to date is only indirectly related to image identification, location, and mensuration, or interpretation per se. Generalization from a combination of these basic functions. The these more theoretical studies to the prob­ task may also include functional and traffic lems of operational interpreters cannot be analyses, damage assessment, and depth made easily. There are several notable ex­ determination to name a few. ceptions, however. On the theoretical side, there is the extension of Tanner and Swet's The manifold problems can be summarized auditory signal detection theory to the prob­ in two broad human factors questions: (1) lems of visual detection (1954). Their basic What are the skills, abilities, and techniques ideas of optimizing performance, taking into necessary to extract information from con­ consideration signal and noise distributional ventional and newer types of imagery? (2) probabilities, along wi th the value of correct How can available human resources be utilized detections and correct rejections and the cost to cope with the ever increasing variety and of false alarms and incomplete identifications, amount of imagery while maintaining ac­ can be applied to the problem of image inter­ ceptable standards of speed, accuracy and pretation. completeness? In order to answer these two broad ques­ In an experiment recently conducted by the tions, an integrated program embracing the Army Personnel Research Office, the effect on following research areas is needed. performance of varying the value of correct 1. Identification of basic human factors in and incorrect identifications was explored. image interpretation Three matched groups of interpreters were 2. Interpreter viewing techniques for near given the same imagery, but each group was real-time and multi-sensor imagery given different information about the relative 3. Team approaches to image interpreta­ importance of identifying all targets and of tion avoiding errors. ApP3.rently emphasis on 4. Image interpreter functions in advanced accuracy, in the situation where there was an equipment configurations ammunition shortage, significantly reduced 5. Techniques for selecting image inter­ the number of erroneous identifications with­ preters out reducing the number of correct identifi­ 6. Training techniques for image inter­ cations made (see Table 1). preters Considerable research in the basic factors Before rapidly summarizing some of the area has been directed toward study of the latest research being conducted or planned in relationship of such physical characteristics these areas, it should be emphasized that the of imagery as image scale, contrast, granu­ research to date represents only a promising larity, and sharpness to the usefulness of beginning. And, unfortunately, it is basically imagery for the extraction of information. an uncoordinated beginning. There has not Research in this area has suffered from the been sufficient communication or integration limited range of physical variables studied, of research findings and plans to enable re­ the artificiality of the imagery employed, the searchers to build upon the efforts of each lack of experimental controls utilized, and other. This is not surprising. The problems inadequate performance criteria. Recently, are complex and researchers tend to bite off however, the Cornell Aeronautical Labora­ what little they can chew in a reasonable tories and the Military Products Group of period of time. However, if substantial Minneapolis-Honeywell, both under contract progress is to be made, not only must duplica­ to the Air Force's Rome Air Development tion of effort be avoided, but problem areas Center. have initiated more comprehensive 980 PHOTOGRAMMETRIC ENGINEERING 100 Research in the second major human fac­ tors research area, in terpreter viewing tech­ 80 Probability niques for near real-time and multi-sensor of Frame imagery, should be closely tied in with re­ Selection Go search in the basic factors area. True, unique problems arise in the rapid scanning and detection of objects in near real-time and in 20 the integrative viewing of imagery from several sensors. But many of the basic prob­ 15 lems involved are identical to those present in Number of Targets Per Print the interpretation of photographic imagery under conventional conditions. Significant FIG. 1. Screening efficiency within research in near real-time viewing has recently 10 seconds by target density. begun. The Applied Psychology Corporation, under contract to the Rome Air Development studies in this area. The principal objective of Center, is studying the effects on interpreter the Cornell study is to determine how physical accuracy and completeness of viewing times, image variables can best be combined into a measured in seconds, and various work-rest summarv measure which maximally predicts cycles. The effects will be studied across interpreter performance. The goal of the imagery varying in scale, resolution, contrast, Minneapolis-Honeywell study is to deter­ and target density and detail. Preliminary mine the accuracy and speed with which results indicate that the number of targets selected objects of mili tary significance can be and features correctly identified increase as a identified at various levels of ground resolu­ function of logarithmic increases in the tion and contrast. These studies are about amount of viewing