Improved TransT Target Tracking Algorithm Based on Hybrid Attention

Abstract

The TransT method pioneered the introduction of the attention mechanism into the target tracking field, but there are still shortcomings in stability. To improve the stability of the algorithm, we propose an improved feature fusion algorithm for visual image target tracking, and we adopt the LKA attention mechanism to strengthen the local attention to the target to ensure the tracker achieves both global and local attention. The tracking of visual image targets is achieved. The experimental results show that the proposed method has a short execution time and high tracking accuracy.