中文摘要：

为降低遮挡对目标跟踪性能的影响,提出了一种基于自适应更新时空方向能量的目标跟踪算法.首先依据目标外观模型进行初步跟踪,计算目标平均运动矢量;然后,求取运动目标的时空方向能量特征,构建运动模型;依据运动模型和状态机检测目标状态,生成遮挡掩膜;最后,对不同状态和遮挡情况的目标采用不同的参数自适应更新其外观和运动模型.实验采用国际通用的CAVIAR和York两个公共测试数据集,并用平均跟踪误差和多目标跟踪精确度两个指标评测了跟踪性能.实验结果表明该方法的目标跟踪性能好,尤其是对目标遮挡的鲁棒性强.

英文摘要：

To reduce the impact of occlusion on target tracking performance,a target tracking algorithm based on spatiotem?poral oriented?energy of adaptive update is proposed. With the algorithm,the moving target is tracked in preliminary according to its appearance model to compute the average motion vector,and then obtain the spatiotemporal oriented?energy feature of the target and construct the motion model. After that the target state is detected according to the motion model and state machine to generate the occlusion mask. For the target with different status and occlusion conditions,the appearance model and motion model of the target are self?adaptively updated by means of different parameters. In the experiment,two testing datasets(CAVIAR and York) commonly used in the world were adopted,and the indexes of average tracking error and multi?object tracking precision were employed to evaluate the tracking performance. The experimental results show that the proposed method has good target tracking performance,especially for strong robustness to target occlusion.