基于混合相关滤波信息融合再检测的目标跟踪算法

为了减小在目标跟踪过程中目标形变和复杂背景变化对跟踪效果的影响,提出一种基于混合相关滤波信息融合再检测的目标跟踪算法。首先,利用相关滤波算法提取到目标的方向梯度直方图HoG特征,利用颜色模板得到目标的颜色特征,计算两个模板的采样得分;其次,再将两者的特征信息用线性组合的形式进行特征信息融合确定目标位置,跟踪过程中,根据设定的阈值条件选择两个模板采样较大的得分再检测目标的位置;最后,输出所有帧目标位置的结果。与其他的算法进行比较,该算法在应对目标形变和背景杂波方面有较好的跟踪效果。

Target Tracking Algorithm Based on Hybrid Correlation Filtering and Information Fusion Redetection

In order to reduce the influence of target deformation and complex background changes on the tracking effect during target tracking, a target tracking algorithm based on hybrid correlation filtering information fusion redetection is proposed.Firstly, the correlation filtering algorithm is used to extract the HoG feature of the target gradient histogram, the color template is used to obtain the color feature of the target, and the sampling score of the two templates are calculated.Secondly, the two feature information is combined in a linear combination to determine the target position, during the tracking process, two templates are selected to sample a larger score according to the set threshold condition and then redetect the position of the target.Finally, it outputs results for all the frame target positions.Compared with other excellent algorithms, the algorithm has better tracking effect on dealing with target deformation and background clutter.