加权模型下的相似匹配跟踪方法

针对传统跟踪方法易受相似物遮挡而导致丢失目标问题，提出一种加权模型下的相似匹配跟踪方法。首先，将目标区域分割成局部特征块，并为其分配权重，建立带权局部特征块组成的外观模型；然后，利用目标的颜色、位置特征进行相似性匹配，为了避免复杂背景干扰，在匹配前划分前景区域，从而实现较准确跟踪；最后，提出一种遮挡决策模型更新机制，通过对目标发生严重遮挡进行判定，保证模型的匹配鲁棒性。实验结果表明，利用加权模型以及多特征相似匹配，使得该方法能够得到较高的跟踪准确率，平均误差仅为13.21，跟踪重叠率为0.71。

Similarity matching tracking method under weighted model

Focusing on the problem that the traditional tracking method is susceptible to occlusion of similar objects, this paper proposed a similarity matching tracking method which based on weighted model. According to weighted theory, it divided the target region into local feature blocks, assigned different weights to them, therefore it established the local weighted model of target. Then it used the target color and location characteristics to calculate the similarity between the test images and the target model. In order to avoid the interference of complex background, the method divided the foreground area before model matching, which could get more accurate matching results. Finally, it proposed a novel method of updating occlusion decision model, which ensured the robustness of the model matching by judging the serious occlusion of the target. The experimental results show that the proposed method can achieve high tracking accuracy by using weighted model and multi-feature similarity matching. And the average center error(ACE) of this tracker is only 13.21. The overlap rate is 0.71.