关键词: 粒子滤波, 约束知识, IP-MCMC抽样, PN学习

Abstract: Particle filter algorithm for target tracking with interference may occur in the problem such as lower particle diversity and reduced precision. In allusion to this instance, a novel tracking method based on constraint knowledge is proposed. This method improves the precision of the particle prediction using constraint knowledge. The problem of the particle degeneration is effectively solved by improving the particle diversity with the parallel IP-MCMC method. On this basis, the proposed method realizes the online study algorithm using PN learning, which is used to update the sample distribution of particles and the training samples of the detector. The accuracy and adaptability of the tracking method under complex background is effectively improved. Experimental results show that the proposed method has good effect under the situation of various interference（e.g., shade, deformation, illumination change）.

Key words: particle filter, constraint knowledge, IP-MCMC sampling, PN learning