基于Kalman滤波和区域匹配的视频多目标跟踪
发布时间：2019-04-19 15:41
【摘要】：提出了1个基于Kalman滤波和局部区域匹配的足球球员检测算法。首先利用改进的主颜色分割提取场地,使得预设参数个数降为1个,增加了算法的适用性;其次对球员进行运动跟踪,根据Kalman滤波器预测的位置设置局部搜索区域,采用局部构图的方法获得球员最佳匹配,并在球员遮挡情况下对不同队球员和裁判间的分离采用颜色直方图匹配和质心判定,对同队队员遮挡辅助相对位置及运动轨迹跟踪。实验证明,该算法对于多目标在场景中的遮挡、分离、消失及出现均具有很好的处理效果。
[Abstract]:A soccer player detection algorithm based on Kalman filtering and local area matching is proposed. Firstly, the improved main color segmentation is used to extract the site, which reduces the number of preset parameters to 1, which increases the applicability of the algorithm. Secondly, the players are tracked, the local search area is set according to the position predicted by the Kalman filter, and the best match of players is obtained by the method of local composition. The color histogram matching and centroid decision are used to separate the players and referees under the condition of player occlusion, and the auxiliary relative position and motion track of the same team members are occluded by using the color histogram matching and centroid decision. The experimental results show that the algorithm has a good effect on the occlusion, separation, disappearance and appearance of multi-targets in the scene.
【作者单位】： 河北工业大学电子信息工程学院;河北工业大学计算机科学与软件学院;
【基金】：天津市科技计划项目(14RCGFGX00846) 河北省自然科学基金面上项目(F2015202239)
【分类号】：TN713;TP391.41
，
本文编号：2461076
[Abstract]:A soccer player detection algorithm based on Kalman filtering and local area matching is proposed. Firstly, the improved main color segmentation is used to extract the site, which reduces the number of preset parameters to 1, which increases the applicability of the algorithm. Secondly, the players are tracked, the local search area is set according to the position predicted by the Kalman filter, and the best match of players is obtained by the method of local composition. The color histogram matching and centroid decision are used to separate the players and referees under the condition of player occlusion, and the auxiliary relative position and motion track of the same team members are occluded by using the color histogram matching and centroid decision. The experimental results show that the algorithm has a good effect on the occlusion, separation, disappearance and appearance of multi-targets in the scene.
【作者单位】： 河北工业大学电子信息工程学院;河北工业大学计算机科学与软件学院;
【基金】：天津市科技计划项目(14RCGFGX00846) 河北省自然科学基金面上项目(F2015202239)
【分类号】：TN713;TP391.41
，
本文编号：2461076
本文链接：https://www.wllwen.com/kejilunwen/dianzigongchenglunwen/2461076.html
