In order to solve the problems of weak anti-interference ability and poor robustness of the current target tracking algorithm based on Siamese network，a tracking algorithm which uses channel and spatial attention fusion based on SiamCAR is proposed. The improved efficient channel attention module and spatial attention module are cascaded between the feature extraction subnetwork and the classification regression subnetwork，so as to strengthen the network’s attention to the important channel features and location features in the response map after cross-correlation，and suppress the unimportant feature information. On OTB100，the success rate and accuracy of the proposed algorithm are improved by 3.1% and 2.8% compared with SiamCAR under the challenge of background clutter，respectively. And the robustness and expected average overlap rate of the proposed algorithm are increased by 4.9% and 2.2% respectively compared with SiamCAR in VOT2018. Experimental results show that the proposed algorithm enhances the robustness of the tracker and improves the tracking effect of the tracker in complex scenes.