Improved DP-TBD methods based on multiple hypothesis testing for target early detection

The problem of target detection via dynamic programming based track before detect (DP-TBD) is considered in this paper. DP-TBD is a batch processing method that jointly processes the multiple consecutive frames of measurements in the processing window (PW). Due to this, when target-originated measurements becomes available after the PW beginning, DP-TBD has a detection latency problem, since the merit function of the target state is corrupted by the noise only frames in the PW. To solve this, two improved DP-TBD methods, traversal searching DP-TBD (TS-DP-TBD) and Bayes iteration DP-TBD (BI-DP-TBD), are proposed. Both the two methods use the Multiple hypothesis testing to adaptively adjust the number of jointly processed frames, so as to eliminate the interference of noise only frames. Among them, TS-DP-TBD is the optimal solution but leads to a heavy computational load. BI-DP-TBD is an approximation of TS-DP-TBD, achieving remarkable computational reduction with negligible performance loss. Numerical results show that both TS-DP-TBD and BI-DP-TBD can efficiently solve the detection latency problem and provide superior detection performance.

I. Introduction

Conventional tracking algorithms [1] make detection on each measurement frame, then put the survived measurements into a subsequent tracker. It suffers performance deterioration when the target signal to noise ratio (SNR) is low since the pre-tracking detection only retains the binary results and discards all the rest of information.