Visual object tracking based on visible images becomes less reliable in rainy,foggy weather and the scene with changing lighting conditions.In contrast,infrared images can provide clear imaging in such circumstances by capturing the heat source information of the object.Therefore,by leveraging the complementary advantages of infrared and visible images,robust tracking can be achieved in complex scenes. This article utilizes a multi-scale feature extraction structure in the feature extraction layer to obtain the features of the target in different scale receptive domains,and enhances the feature selection ability through adaptive weight allocation.The polarized attention mechanism is used to enhance the features,thereby achieving effective fusion of dual-mode information for robust tracking.The proposed algorithm is experimentally validated on large infrared and visible light datasets GTOT and RGBT234,with accuracy and success rates of 90.05%,72.7% and 78.7%,55.5%,respectively.The experimental results show that compared with current mainstream algorithms,the algorithm proposed has higher tracking accuracy and success rate.