Target Tracking Based on Improved Particle Filter Algorithm and Camshift Method

The applications of visual tracking in real life are becoming more and more extensive, and attention paid to it is getting higher and higher. The problem of traditional particle filtering is that the recommended distribution probability density generally does not include the target current position state information. The Camshift algorithm introduced in the particle filtering prediction particle phase is proposed, and the current position information is added to the new sampling particle. Using the directional feature, the target edge is extracted by Snake, the observation probability of the directional model feature is calculated. The algorithm greatly reduces the problem of particle degradation and improves the tracking accuracy. And the experimental results show that the algorithm can be used to solve the single-target and multi-target tracking problems with good robustness.