New Feature Extraction And Matching Algorithms For Automatic Target Recognition In Synthetic Aperture Radar

New Feature Extraction and Matching Algorithms for Automatic Target Recognition in Synthetic Aperture Radar

A project to develop an automatic target recognition (ATR) algorithm for synthetic aperture radar (SAR) imagery data, matching an unknown target to one of the known reference targets based on a maximum likelihood estimation procedure. Central to the algorithm is the CLEAN method, which tries to strengthen peak feature image classification importance. The effectiveness of the CLEAN algorithm will be assessed by comparing target recognition accuracy of CLEANed images to those that have not undergone the CLEAN method.

Algorithms for Synthetic Aperture Radar Imagery

Comparison of Different Feature-based and Intensity Signature-based Matching Algorithms for Automatic Target Recognition in Synthetic Aperture Radar

The object of Automatic Target Recognition (ATR) for Synthetic Aperture Radar (SAR) involves comparing extracted target signatures (features) to the statistics of features of all potential targets. Central to this processing paradigm is the search algorithm, which helps assess and optimize the favorable effects of multiple image features on recognition accuracy. The ATR algorithms discussed fall into two categories: feature and intensity-based. The feature-based algorithms create binary images of the edges, corners, gradient and ceiling peaks of the tank. The intensity-based signatures are created using algorithms that extract the tank image, block out the background, normalize the MSTAR (Moving and Stationary Target Acquisition and Recognition) data within the tank region and can be exploited to minimize false classifications. Several scenarios will be explored to determine the effectiveness of using the CLEAN method on the ceiling peak feature extraction method and the validity of using the intensity signatures of the MSTAR tank image.