Heterogeneous Sensor Data Fusion for Target Classification Using Adaptive Distance Function

Abstract

Automatic Target Recognition (ATR) systems are used as decision support systems to classify the potential targets in military applications. These systems are composed of four phases, which are selection of sensors, preprocessing of radar data, feature extraction and selection, and processing of features to classify potential targets. In this study, classification phase of an ATR system having heterogeneous sensors is considered. We propose novel multiple criteria classification methods based on modified Dempster-Shafer theory. Ensemble of classifiers is used as the first step probabilistic classification algorithm. Artificial neural network and support vector machine are employed in the ensemble. Each non-imaginary dataset coming from heterogeneous sensors is classified by both classifiers in the ensemble, and the classification result that has higher accuracy ratio is chosen for each of the sensor. The proposed data fusion algorithms are used to combine the sensors' results to reach the final class of the target. We present extensive computational results that show the merits of the proposed algorithms.