Reinforced Background Aware Correlation Filter with Adaptive Weighting Strategy for RGB-T Tracking

Nowadays, the fusion tracking based on visible and thermal infrared image (called RGB- T) has been developed faster and faster. However, the existing fusion tracking methods cannot have a higher tracking success rate. In order to solve this problem, we propose a reinforced background aware correlation filtering method, which is a strategy of fusion first and then tracking. This method first converts the thermal image into a single-channel image, and then uses the grayscale information to determine the degree of pixel difference between the target and the overall environment. Next, an adaptive weighting decision is made on the two images, and finally the background aware correlation filtering method is applied to the fusion image to realize the fusion tracking. Numerous experiments on the RGBT234 data set show that our proposed fusion tracking method has better performance.