针对目标遮挡、形变、旋转、光照变化及背景干扰等复杂场景下的目标跟踪问题,基于核相关滤波跟踪算法和统计颜色特征的跟踪算法,提出了一种复杂场景下的高置信度更新策略互补跟踪算法。首先,利用高斯拉普拉斯算子和局部二值模式增强目标边缘信息和纹理特征;然后,引入可调高斯窗口函数和基于关键点的尺度估计模型优化算法;最后,利用响应峰值与跟踪框的交并率设计了一种高置信度更新策略,以自适应更新模板。实验结果表明,在OTB2013数据集上本算法的精确度和成功率分别为88.3%和72.4%。

This paper proposes a complementary tracking algorithm with high-confidence updating strategy to address target tracking problems in complex scenes such as target occlusion, deformation, rotation, illumination changes, and background interference. The algorithm is based on the core-related filter-tracking algorithm and the statistical color feature-tracking algorithm. First, the Laplacian of Gaussian operator and local binary mode are used to enhance the edge information and texture features of the target. Then, the tunable Gaussian window function and scale estimation model based on the key points optimization algorithm are introduced. Finally, the response peak value and a high-confidence updating strategy are designed for the merged rate of the tracking frame to adaptively updating the template. Experimental results show that the precision and success rate of the algorithm on the OTB2013 data set are 88.3% and 72.4%, respectively.