Moving object visual attention model particle filter based detection method

The present invention discloses a method for detecting moving target particle filter based visual attention model, based on Bayesian estimation principle firstly, the fusion construct bidirectional attention model particle filter; particle filter and then the bidirectional frame fusion attention model as the basis, moving target color attention and attention B-U, respectively, and T-D input focus, to change the particle distribution is calculated by particle weight, significant attention FIG formed, and finally determines a motion target position. The method of the present invention, attention fusion time and space, so that more accurate calculation of the motion attention; fusion bottom-up and top-down attention, simple and effective simulation of the human visual attention forming process; for complex global motion scenes, improve the effectiveness and accuracy of the moving target detection.