基于Contourlet和MeanShift的交通视频车辆跟踪算法

Video vehicle tracking algorithm based on Contourlet and MeanShift

提出一种基于Contourlet直方图的MeanShift交通视频车辆目标跟踪算法.首先,利用Contourlet变换提取交通视频中感兴趣区域下跟踪目标的纹理特征和轮廓,并用其直方图统计跟踪目标的纹理特征;然后,通过Kalman滤波技术来进行跟踪目标轨迹的相关性函数计算,并将计算结果迭代到MeanShift跟踪算法中选取最优估计值,进而实现对交通视频车辆在复杂场景中的精确定位.通过与传统的MeanShift跟踪算法以及基于Kalman滤波的MeanShift跟踪算法等比较表明,此算法不仅能够对复杂场景中的视频车辆进行有效跟踪,同时还具有较好的稳定性和抗干扰性.

Based on Contourlet Histogram, the paper proposed a new traffic video tracking algorithm of MeanShift. First of all, using Contourlet transform to achieve the textural feature and contour of the tracking object in the region of interest from traffic video,and using histogram to count tracking ob- ject＇s textural characteristics. Then, calculate the correlation function of tracking object curve with Kalman filter. Finally,iterate the computing result to MeanShift tracking algorithm to select the opti- mal estimate, and then realize the accurate location of traffic video vehicles in complex scenes. By MeanShift of traditional tracking algorithm with MeanShift tracking algorithm based on Kalman filter comparing, we find shows that this algorithm not only can effectively track the video vehicle in com- plex scenes, but also has good stability and anti-jamming.