Target Detection and Classification Algorithm Using Seismic Sensor and Pulse Doppler Radar

Abstract

This thesis research achieved its primary scientific objective to perform robust, automatic detection and classification of moving targets using seismic sensor and pulse doppler radar on a sta-tionary platform. Detection and classification algorithm using seismic sensor and pulse doppler ra-dar signal is a problem of current interest. The purpose is to detect and classify the moving target without human aid. The target classes were included human running, human walking and animal. Evaluation of theory on realistic experimental data is vital to the advancement of knowledge. The data can be used to rigorously evaluate new classification, detection, and feature selection algo-rithms. Computer simulation analysis also plays a crucial role in theoretical development. The expe-rimental data collected by this thesis research can be utilized to improve the accuracy of computer models.This thesis contributed a novel set of high-performance seismic and doppler based features. The Fisher used for selecting a feature set. In addition, the feature set included both statistical and the linear predictive coding (LPC) residual energy feature. The selected feature set was shown to perform well on the seismic sensor and doppler radar based target classification problem. The design and detailed analysis of target classification algorithms based on support vector machine (SVM) and binary tree architecture (BTA) classifiers were designed to accomplish high-performance target classification. The importance of both classifier selection and feature selection was analyzed in detail. In order to process the complex-valued signals from pulse doppler radar, complex-valued SVM classifier was derived. Complex-valued SVM classifier processes the complex-valued signals measured by PDR to identify moving targets from the background.