An algorithm for detecting groups of targets

In support of the ARPA-sponsored Critical Mobile Targets program, Lincoln Laboratory has developed a complete, end-to-end, automatic target recognition (ATR) system using very high resolution (1 ft×1 ft), fully polarimetric, Ka-band SAR data. Lincoln Laboratory has also developed ATR systems for reduced resolution (1 m×1 m) and single polarization data; performance comparisons (1 ft×1 ft versus 1 m×1 m and fully polarimetric versus single polarization) have been performed. There has been interest in the use of automatic target cueing (ATC) followed by ATR. In military operations, targets are often deployed in well-defined groups, such as tank companies. It is possible to exploit this information by developing algorithms to detect such groups of targets. Cueing algorithms identify potential target groups; then these areas can be revisited with a higher resolution sensor or these areas can be processed with more sophisticated algorithms (or both). This paper presents results from a joint ATC/ATR system that uses single polarization (HH) 1 m×1 m resolution data for both ATC and ATR; descriptions of the algorithms in the system are given and the performance is compared to the (1 ft×1 ft) very accurate resolution system run on the identical data set. The data used in this study consisted of 231 km² of 20 degree depression angle stripmap clutter from several areas in the northern United States; these data had varying amounts of cultural content from benign to severe