SAR Imaging via Modern 2-D Spectral Estimation Methods. Volume 2: Impact on Automatic Target Recognition.
Final rept. May 94-Jun 95,
ENVIRONMENTAL RESEARCH INST OF MICHIGAN ANN ARBOR
Pagination or Media Count:
This report discusses the results of a study to evaluate the impact of several promising adaptive image formation processing AIFP methods on the performance of ATR algorithms for fine resolution SAR images of ground order of battle GOB targets. The AIFP methods are an extension of modern 2D spectral estimation techniques. Relative to traditional Fourier image formation processing, they offer the potential for improved image resolution, enhanced target to clutter contrast, and reduced speckle levels. Three specific AIFP methods were considered adaptive sidelobe removal, spatially variant apodization, and minimum variance. ATR performance was analysed using a correlation based target classification algorithm that is representative of state of the art SAR template matching approaches. The results showed that SVA image quality is significantly better than Taylor weighted Fourier image quality for the identical phase history data set. The results also showed that SVA imagery offers the potential for improved target classification performance relative to Taylor weighted Fourier imagery. The simple correlation based target classification algorithm performed well in separating tanks of different classes from each other and from other confusion target classes.
- Active and Passive Radar Detection and Equipment