The Research for Pedestrian Tracking Algorithm with Mutil-feature Fusion

In order to overcome occlusion and to accurately track objects,the thesis proposed a kind of multi-featured tracking algorithm,which is based on the Nearest Neighbor. The color and geometrical features are the most visualized and they can be used to track objects accurately.When the occlusion appeared, according to the color and geometrical features of the target object will be unable to track the target. Harris corner can be utilized to deal with some occlusion. Therefore, these three features, namely color,geometry and Harris, can be mixed together to overcome occlusion. However, the mixture of many features will reduce the timeliness of the system. The thesis decided the priority of the object matching by using nearest neighbor and overcome the bad effects of multifeatures on timeliness. As experimental results show, the algorithm proposed in this paper has good tracking accuracy and robustness to the target’s deformation and block, and overcome the problem of poor timeliness which bring with feature fusion.