基于MATLAB的相关滤波跟踪算法仿真分析

摘要：基于相关滤波的目标跟踪算法是一种常见的视觉跟踪方法，它利用目标的特征信息进行跟踪；在跟踪过程中通过计算目标模板与当前帧图像中候选区域之间的相关性来判断目标的位置；通过介绍首个将相关滤波理念与目标跟踪技术相结合的MOSSE算法，引入了三种基于此算法的改进相关滤波跟踪算法：KCF算法、DSST算法以及BACF算法；并基于视频跟踪基准OTB100数据集在MATLAB平台进行仿真实验，一次性评估标准下，BACF算法的平均成功率与平均精确度分别为最高的64.5%与80.4%，空间鲁棒性评估标准下，BACF算法的平均成功率与平均精确度分别为最高的58.2%与78.6%，时间鲁棒性评估标准下，BACF算法的平均成功率与平均精确度分别为最高的65.8%与85.1%，因此BACF算法的跟踪性能最佳，而KCF算法实现了最高的154.36帧率的跟踪速度。

Simulation Analysis of Correlation Filtering Tracking Algorithm Based on MATLAB

ABSTRACT:The target tracking algorithm based on correlation filtering is a common visual tracking method, which uses the feature information of the target to track; In the tracking process, the target position is determined by calculating the correlation between the target template and the candidate area in the current frame image; Through the introduction of the first MOSSE algorithm combining the concept of correlation filtering with target tracking technology, three improved correlation filtering tracking algorithms based on this algorithm are introduced: KCF algorithm, DSST algorithm and BACF algorithm; And conduct simulation experiments on MATLAB platform based on the OTB100 data set; Under the one-time evaluation standard, the average success rate and average accuracy of BACF algorithm are the highest 64.5% and 80.4% respectively; Under the spatial robustness evaluation standard, the average success rate and average accuracy of BACF algorithm are the highest 58.2% and 78.6% respectively; Under the time robustness evaluation standard, the average success rate and average accuracy of BACF algorithm are the highest 65.8% and 85.1% respectively; Therefore, BACF algorithm has the best tracking performance, while KCF algorithm achieves the highest tracking speed of 154.36 frame rate.