Aiming at the problem that the output of correlation filter is invalid due to the change of brightness in target tracking process, this paper proposed a correlation filter tracking method based on video brightness optimization. This method used mutation and population diversification strategy to enhance the global optimization ability of particle swarm optimization algorithm. By establishing the mapping relationship between the average brightness of the video sequence and the variance of the output result of the correlation filter, it obtained the best expression of the output variance of the correlation filter under different brightness conditions. At the same time, it determined the update speed of the correlation filter template by calculating the similarity between adjacent search frames, which reduced the probability of tracking failure and the amount of calculation. Experiments show that the optimized tracking framework has better target adaptability and robustness under different brightness and background conditions.

[2019-12-05] Printed Article /doc/2018.08.0600/v2