Terminative joint sequential object detection and tracking based on fused test statistics

Joint object detection and tracking is a powerful approach to significantly improve the detection of extremely weak targets or phenomena in surveillance systems. Since the Kalman filter is an optimal estimator for object tracking problems under certain conditions and the Wald's sequential probability ratio test (SPRT) requires fewer samples in average than the fixed-sample-size procedure when solving object detection problems, it is beneficial to apply the Kalman filter and the Wald's SPRT to design joint object detection and tracking algorithm. However, the Wald's SPRT cannot be rigorously proved to eventually terminate if the observations are dependent. In this paper, a terminative joint sequential detection and tracking approach is proposed by fusing two test statistics: one is derived in our previous work, and the other is based on independent observations obtained by linearly combining a group of adjacent measurements. The proposed approach takes advantage of both statistics in that it is guaranteed to terminate and it requires on the average a small number of measurements. Numerical results show that the average sample number required by the proposed approach is very small under low signal-to-noise ratio conditions and the actual probabilities of errors are smaller than the nominal probabilities of errors.