Investigation of Ground Target Detection Methods in Fully Polarimetric Wide AngleSynthetic Aperture Radar Images

摘要

Target detection is a high priority of the Air Force for the purpose ofreconnaissance and bombardment. This research investigates and develops methods to distinguish ground targets from clutter (i.e. foliage, landscape etc.) in Wide Angle Synthetic Aperture Radar (WASAR) images. WASAR uses multiple aspect angle SAR images of the same target scene. The WASAR data was generated from a pre-release software package (XPATCH-ES) provided by the sponsor (WL-AARA). A statistical analysis and feature extraction is performed on the XPATCH-ES data. Polarimetric and wide angle covariance matrices are estimated and analyzed. From an analysis of the wide angle covariance matrix it is shown that natural clutter has in general a uniform radar return for changing aspect angles, whereas the radar return for a target varies. Based on this analysis, two new wide angle algorithms, the WASAR Whitening Filter and the Adaptive WASAR Whitening Filter (AWWF) are developed. The target detection performance of polarimetric and multi aspect angle image combining algorithms are quantified using Receiver Operating Characteristic curves and target to clutter ratios. It is shown that wide angle processing provides superior target detection performance over polarimetric processing. Combinations of wide angle and polarimetric algorithms were used to achieve a 13.7 dB processing gain in target to clutter ratio when compared to unprocessed images of the target scene. This represents a significant improvement in target detection capabilities.