Particle filter tracking based on visual saliency feature

To address the problem of moving object tracking in complicated scenes, a particle filter tracking algorithm based on visual saliency feature was presented. The algorithm detects the object in the image with the saliency detection algorithm to get saliency maps. Target states are predicted using the second-order autoregressve model, and the final saliency map is obtained with the center-strengthening and edge-weakening mechanism. The saliency feature is extracted according to the phenomenon that in the saliency map pixel value is greater when the pixel is in the target area, and is then fused with the color feature adaptively to complete tracking. Experimental results show that the algorithm can effectively deal with the situation when illumination and appearance change.