The Measurement of Stimulus Complexity in High-Resolution Sensor Imagery

摘要

A Stimulus Complexity Analyzer was built and used to obtain measures of overall background complexity on several types of high-resolution reconnaissance imagery. These measures were then compared to target recognition performance in a series of studies using Side Looking Airborne Radar (SLAR), infrared (IR), and aerial photographic imageries. Significant relationships between a number of analyzer measures and recognition performance were obtained. For the photo and IR imagery backgrounds studied a perfect curvilinear relationship was obtained between analyzer measures of total object count and recognition performance. Inverse relationships were obtained between target recognition time and mean object size and object size variance for the SLAR, IR, and photo backgrounds. These relationships were more obvious with the photo and IR imageries than with the SLAR imagery. A technique for embedding targets in different backgrounds was developed and used. It allowed the effects on performance of target differences, background differences, and target-by-background interaction to be studied. (Author)