Hierarchical foveal algorithm development for ATR

Period of Performance: 07/05/1995 - 07/05/1997


Phase 2 SBIR

Recipient Firm

Amherst Systems, Inc.
30 Wilson Road
Buffalo, NY 14221
Principal Investigator

Research Topics


Foveal active vision features imaging sensors and signal processing with graded acuity coupled with context sensitive gaze control, analogous to that prevalent throughout vertebrate vision. Foveal vision operates more efficiently in dynamic scenarios than uniform acuity vision because resolution is treated as a dynamically allocatable resource. Wide field-of-view (FOV) and central high acuity are simultaneously supported while minimizing data to that which is relevant to the task. The proposed program will develop a Hierarchical Foveal Machine Vision (HFMV) ATR capability consisting of detection, recognition, and gaze control algorithms operating in a closed-loop software simulation environment. The performance of HFMV ATR will be evaluated and compared against the performance of uniform acuity ATR. To ensure fairness in the comparison, both systems will use the same sensor data rate and ATR techniques. Specifically, the retinotopology of the HFMV ATR algorithms will be parametic so as to support a range of acuity profile gradients; a flat profile will be that identified in Phase I as optimum for ATR. The simulations will use second generation measured FLIR imagery to test foveal recognition with realistic data, and large frame synthic imagery to test peripheral detection and saccadic gaze control.