基于快速运动场景下的目标跟踪改进算法

为解决目标快速运动时跟踪算法出现目标丢失和跟踪精度大幅度下降等问题,在现有的Autotrack算法基础上对其进行改进,提出了一种基于快速运动场景下的目标跟踪算法。引入空间正则权重项w对距离目标中心比较远的样本进行相应的惩罚,调整原本的全局响应变化量并将其作为时间正则项。将空间正则项和时间正则项相结合,并引入目标函数中进行优化。在公开数据集OTB-2013(online object tracking:a benchmark)上对改进后的目标跟踪算法进行实验验证和比较。实验结果表明,改进后的目标跟踪算法在目标快速运动场景下的准确率和成功率分别为76.5%和73.1%,在综合评分上的准确率和成功率分别为82.8%和61.1%。

Improved target tracking algorithm based on fast moving scene

In order to solve the problems of target loss and a significant reduction in tracking accuracy when the target is moving fast, the existing Autotrack algorithm is improved and a tracking algorithm based on fast moving scenarios is proposed. The spatial regular weight term w is introduced to penalize samples that are far away from the target center, and the original global response variation is retained and adjusted as the time regular term. The combination of the two is introduced into the objective function for optimization. The experimental validation and comparison on the public dataset OTB-2013 (Online object tracking: A benchmark) shows that the improved tracking algorithm has significantly improved the corresponding accuracy and success rate in the fast target movement scenario. 82.8% and 61.1% are achieved on the composite score. 76.5% and 73.1% are achieved on the accuracy and success scores for the fast motion scenario, respectively.