基于上下文感知与自适应响应融合的相关滤波跟踪算法

针对当前相关滤波跟踪算法在抗背景干扰、响应融合方式以及模型更新策略上的不足,提出一种基于上下文感知与自适应响应融合的相关滤波跟踪算法.通过引入上下文感知技术,提高算法在背景杂波及遮挡等跟踪场景下的鲁棒性;通过研究HOG特征和颜色直方图特征二者响应图和响应值的特点,提出一种自适应响应融合方法,提升融合响应图的可靠性;在模型更新方面,采用了高置信度模型更新策略来减轻传统模型更新策略中模型污染及跟踪漂移的问题.实验结果表明,本文算法在OTB50数据集上达到了74.7%的跟踪精度,跟踪成功率为54.8%,均优于对比的主流相关滤波跟踪算法,并且在背景杂波、光照变化、遮挡、运动模糊等复杂跟踪场景中具有较好的跟踪精度与鲁棒性.

Correlation Filter Tracking Algorithm Based on Context-aware and Adaptive Response Fusion

Aiming at the shortcomings of the current correlation filter tracking algorithm in background clutters,response fusion method and model updating strategy,this paper proposes a correlation filter tracking algorithm based on context-aware and adaptive response fusion. By introducing context-aware technology,the robustness of the algorithm in tracking scenes such as background clutters and occlusion is improved. By studying the characteristics of the response map and response value of the HOG feature and the color histogram feature,an adaptive response fusion method is proposed to improve the reliability of the fusion response map. In terms of model updating,a high-confidence model updating strategy is adopted to alleviate the problems of model pollution and tracking drift in traditional model updating strategies. Experimental results demonstrate that the proposed algorithm achieves 74.7% tracking precision on the OTB50 dataset,and the tracking success rate is 54.8%,both of which are better than the mainstream correlation filter tracking algorithms compared,and it has better tracking precision and robustness in complex tracking scenes such as background clutters,illumination variation,occlusion,motion blur and so on.