Multiscale models for target detection and background discrimination in synthetic aperture radar imagery

摘要

Multiscale models can be used to capture the scale-dependent behavior of the statistics in radar imagery. This behavior is expected to be different for natural background compared to objects of interest such as vehicles. We demonstrate that multiscale autoregressive models can discriminate between samples of these two major classes extracted from 1.5-m-resolution radar imagery. We also show that it is possible to discriminate between two types of natural background in SAR imagery, `grassland' and `woodland,' using multiscale models. This latter result could be exploited in adaptive algorithms for automated target detection.