Contourlet detection and feature extraction for automatic target recognition

This research presents a contourlet based detection and feature extraction method for underwater targets. The method operates on Side Scan Sonar (SSS) images and is designed to automatically detect and generate target features for classification. Kernel based classifiers are used to determine the best boundary for separating targets and clutter. A statistically significant target data set is generated by embedding additional synthetic targets into SSS data collected during sea tests. Feature trade off studies show an improvement in classification results with the addition of directional based features.