一种基于高斯概率模型的多人跟踪算法

Multiple Pedestrians Tracking Based on Gaussian Probability Model

针对视频监控中的多人跟踪问题,提出了一种基于高斯概率模型的算法。基于目标颜色的统计特征,采用改进的K均值方法,将目标区域按颜色信息聚类,并根据聚类结果对目标区域分块,然后用高斯模型对各分块分别进行建模。同时,对目标的位置信息也进行高斯建模。通过计算待检测目标与模型之间颜色和位置的最大联合概率值来实现跟踪。利用前后帧中目标的位置信息及颜色信息,能克服目标遮挡后因信息的丢失而跟踪失败的情况。实验结果表明,该算法具有较强的鲁棒性,能有效实现多人的跟踪。

To deal with the multiple pedestrians tracking problem in video surveillance, an algorithm based on Gaussian probability model was proposed. After foreground objects were extracted by adaptive background mixture models, every object was segmented to blocks according to the grouping results of an improved k-means algorithm,which was based on the statistical characteristics of color information. The distribution of color information in each block and the position of the target were described by two Gaussian functions respectively. After the maximum joint probability of color and position information had been obtained, the tracking procedure was finished. Combined the color information and position information together, the occlusion regions could be grouped and distinguished very well, which made the algorithm more robust. Experimental results demonstrated the effectiveness and robustness of the proposed algorithm.