Detection of targets using distributed multi-modal sensors with correlated observations

Multiple unattended ground sensors (UGSs) equipped with seismic, passive infrared (PIR) and ultrasonic sensors are used to detect and track people for better situational awareness. The majority of the false alarms are caused by animals. We fuse the detections of individual sensors at each node using algorithms based on the Neyman-Pearson criteria to achieve the required false alarm rate with the assumption that the sensor observations are independent. However, the sensors are observing the same phenomenon and hence the observations are not independent. In this paper, we explore the joint probability distributions between the sensors using copulas to improve the detection statistics. We identify several copula functions suitable for fusing the data in order to improve the detection statistics.