Abstract:

As the current sparse representation tracking algorithms are of low real- time performance,this paper puts forward a tracking algorithm of the block sparse representation with background information.In the algorithm,first,candidate targets are selected through a fast algorithm in the framework of particle filtering.Then,images are divid-ed into blocks with different weights to reduce the disturbance of occlusions and noises.Finally,the background in-formation is fused into the template dictionary,and the sparse solution of the candidate targets is calculated through the collaborative representation of the targets and the background.The proposed algorithm is compared with the oth-er two algorithms by using four videos on the Matlab platform.The results show that the proposed algorithm is of a higher calculation speed and a stronger robustness.

Key words: object tracking, sparse representation, template dictionary, particle filtering