CDC00-REG1373 Clustering-Based Maximum Likelihood Estimation: Application to Sensor Fusion in UXO detection

Abstract: This paper describes our approach of applying clustering techniques in the detection of UXOs (Unexploded Ordances). The clustering algorithms help us to integrate information from various sensors. Because of the specific characteristic of our application, a new MLE (maximum likelihood estimation) method is designed. The MLE is integrated into the clustering algorithm whose inputs are the extracted features of targets. This approach is tested on one of the MTADS (Multi-Sensor Towed Array Detection System) data. Good performance has been achieved. Finally, it should be pointed out that further improvements could be achieved when better feature extractions become available.