Characterized by a combined multi-level particle filter tracking method

The present invention discloses a hierarchical particle filter tracking method for a multi-joint feature, comprising the steps of: acquiring successive images of the target frame, and the selected initial target region and the initial set of particles in the first frame image; starting from the second frame: in the center point of the target region of the second image frame k-1 k-th frame as a search area center point, corner detection in the search area to obtain a corner point; and create the k-1 frames of the target color and texture joint histogram feature; the set of particles by known motion model at time k; by corner detection and feature points k-th frame and the k-1 frame image matching, determining a weight value at time k-order particles; the joint histogram of the target color and texture features likelihood function is established, the first-order and combined weights obtained by multiplying the likelihood function of second order, and the weight normalized, as the particle weight at time k; k time according to the particle weight and set the value of the particles, the target area acquired at time k; repeat the process until the completion of the particle filter.