Tracking Before Detection Based On Dynamic Programming

The detection and tracking of weak targets is one of the hot-spot aspects in themodern radar technology, and track-before-detect (TBD) is an useful method of weaktargets detection and tracking. TBD directly uses no threshold or low thresholdmeasurements of sensors for adequate information utilization. It gains the increasing ofsignal to noise ratio(SNR) by temporal accumulation of measurements and realizes thedetection and tracking of weak targets simultaneously.This paper mainly studies the dynamic programming(DP)-based TBD algorithm.Its main content comprises:1. The traditional DP-TBD algorithm is introduced for the detection and trackingof weak point target which is in Gaussian noise background and has constantvelocity. The algorithm integrates the measurements along possible targettrajectories, returning as possible targets those trajectories for which themeasurement sum, or merit function, exceeds a threshold. The approach isbased on the principle that the energy accumulated along the target trajectory isthe largest.2. The improved DP-TBD algorithm is proposed. An idea of weighted direction isintroduced to decrease the range of energy diffusion in dynamic programmingalgorithm. A technique of finite predictive strategy is performed to weaken theimpact of strong noise points on the target trajectories. In addition, it uses themethod of two-step detection thresholds to reduce the amount of thecomputation.3. The detection and tracking of multiple targets is studied. First the extremevalue method based on dynamic programming is introduced. As the extremevalue method can not track two targets whose trajectories intersect, bubblingmethod based on dynamic programming is proposed. The algorithm firstdetects the largest intensity target from the original measurements and removesit from the measurements to get new measurements. Then it detects the largestintensity target from the new measurement and removes it. Repeat these stepsuntil all the targets are detected. The essence of this method is changing thedetection and tracking of multiple targets into several single target detectionand tracking.