5 June 1995 Analytical framework for determining the robustness of linear feature mapping for target detection
Author Affiliations +
Abstract
The incorporation of linear feature mapping into a maximum likelihood ratio test criterion makes possible a novel realizable adaptive detector of a 2D resolved signal with limited priori knowledge of the target signal pattern and the nonstationary clutter. As demonstrated by experiments on actual SAR data, the detection probability of the test depends strongly on the effective generalized signal-to-noise ratio (EGSNR), which is determined by the selected feature mapping and representation. In this paper, an analytical framework is proposed for the linear feature mapping detector (LFMD), within which various linear feature mappings, such as the short time Fourier transform (STFT), the discrete cosine transform, the discrete wavelet transform, and the discrete Gabor transform, etc. can be compared in a systematic way in terms of detection performance. The closed-form solution of the porblem is obtained and tested by using the SRI ultra wideband (SAR) data, which is single polarization with 200MHZ to 400MHz band at 1 meter resolution. The robustness of the LFMD with several linear mappings is compared for classifying multi-oriented targets in actual SAR image. Several different classes of targets including both civilian and noncivilian vehicles imaged at broadside and head-on are used for the study. Also the effect of an inaccurate priori knowledge of the target's significant features on the value of the EGSNR is investigated.
© (1995) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Xiaoli Yu, An Mei Chen, Lawrence E. Hoff, Irving S. Reed, "Analytical framework for determining the robustness of linear feature mapping for target detection", Proc. SPIE 2487, Algorithms for Synthetic Aperture Radar Imagery II, (5 June 1995); doi: 10.1117/12.210862; https://doi.org/10.1117/12.210862
PROCEEDINGS
7 PAGES


SHARE
Back to Top