|
This paper discusses measurement of image interpreter performance as means for assessing techniques associated with proposed improvements in reconnaissance technology (i.e., associated with the collection, display, and processing of reconnaissance imagery). Selected examples of research will be described to illustrate this approach including a discussion of methodology for evaluating the contribution of color imagery. The goals of aerial reconnaissance are oriented toward providing detailed, accurate, and timely intelligence information for a multitude of uses. Attaining these goals is often quite difficult. Even with the advanced sensor systems available today, it must be recognized that the extraction of intelligence information is a complex and time consuming process. In order to increase the accuracy and to reduce the time required for information extraction, it is necessary that we have a knowledge of the present extraction capability, anticipated advances in imagery, along with time requirements. It is extremely important that this knowledge be derived from controlled experimental situations which simulate operational conditions and requirements. Reconnaissance systems normally are improved through a step by step refinement of the sensor's ability to discriminate and record objects in the real-world environment. In some cases, scientific break-throughs make possible significant gains in a sensor's ability to record intelligence information. All of this advancement is lost however, if some efficient and expeditious means is not found to (1) extract, (2) analyze and (3) report the pertinent information to a using command. The pursual of carefully planned, executed and analytic research in reconnaissance can lead to a valuable source of answers for a number of critical questions. To cite an example, in the evaluation of a new sensor system the question of primary importance to the developer and/or user might be whether or not timely detection of freight train traffic at night is possible. In setting up an experimental design addressed to this problem, the first step is to determine the (1) questions, (2) parameters, and (3) interactions, etc. which might be expected. The following type of questions might arise: 1. Can freight trains be detected at night by sensor X? 2. What is the maximum range for such a detection? 3. How does weather affect the sensor/re-cord? 4. What advantages does it offer over conventional techniques? 5. What are the time-lags incurred by various information extraction and technical decisions for using the information? Parameters in such a study might include: (a) Aspect angle (b) Altitude (c) Scale (d) Target/background contrast (e) Special interpreter viewing devices (f) Interpreter background/training (g) Interpreter task (screening versus detailed analysis, etc.) When all of these factors have been determined, they must be combined into an experimental design which will reflect the requirements for image acquisition and realistically
|
