Current automated approaches for

SummaryA system for contextual analysis of tactical scenes has been presented. The componentsof the system are functional, and are currently being trained for optimal performance onthermal imagery of ground vehicles in natural backgrounds. Later this year the componentfunctions are to be integrated into a single context -dependent object recognizer whose per-formance will subsequently be measured and compared with that of a typical context- independ-ent object recognizer. The resulting baseline system represents the first steps towards aninfant AI expert vision system, which will be continually enhanced to incorporate future ad-vanced algorithm developments as they occur.References 1. Gilmore, J.F., and Spiessbach, A.J., "A Model Driven System for Contextual SceneAnalysis," SPIE Application of Digital Image Processing VI, Vol. 432, p. 262, San Diego,California, August 23 -26, 1983. 2. Makoto Nagao, Takashi Matsuyama, and Hisayuki Mori, "Structural Analysis of ComplexAerial Photographs," Sixth International Joint Conference of Artificial Intelligence, Tokyo,Japan, 1979.

[1]  W. A. Perkins,et al.  Area Segmentation of Images Using Edge Points , 1980, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[2]  A. J. Spiessbach,et al.  A Model Driven System for Contextual Scene Analysis , 1984, Optics + Photonics.

[3]  Jr. Kenneth Robert Sloan,et al.  World model driven recognition of natural scenes. , 1977 .

[4]  Azriel Rosenfeld,et al.  Scene analysis using region-based constraint filtering , 1984, Pattern Recognit..

[5]  Allen R. Hanson,et al.  Computer Vision Systems , 1978 .

[6]  Avron Barr,et al.  The Handbook of Artificial Intelligence , 1982 .

[7]  Bir Bhanu,et al.  Segmentation of natural scenes , 1987, Pattern Recognit..

[8]  Takeo Kanade,et al.  A Production System for Region Analysis , 1979, IJCAI.