Abstract:

In Gaussian mixture probability hypothesis density (GM-PHD) filter, the new born targets exist the whole detection domain, and the specific position is hard to define for occurring randomly. Therefore, an improve algorithm, using the received measurements adaptively generate target intensity function to record the surviving target intensity function, is realized. The algorithm can adaptively distinguish the surviving and new born target intensity function to improve the accuracy. Multiple target position tracking simulation and the dolphin whistle real data processing are used to test improved algorithm performance, and optional sub pattern assignment (OSPA) function works as a benchmark. The result shows that the proposed algorithm improves the target number estimation and tracking accuracy. The correct rate of target number estimation is 97%, and the OSPA distance is decreased nearly 30% of the original GM-PHD filter.