摘要
实时的轮廓跟踪算法可以为视频监控系统提供物体的轮廓信息以供对物体类别、物体行为等进行识别。提出一种基于均值漂移和边缘检测的轮廓跟踪算法。方法中,首先利用均值漂移算法跟踪得到目标物体的中心位置,同时用高斯统计模型进行背景更新,从前景图像和背景图像中分别得到具有相同位置和大小的前景矩形区域和背景矩形区域,然后用背景分割的方法得到目标物体区域,再对目标物体区域进行边缘检测就得到了目标物体的轮廓,进而实现了对目标物体的轮廓跟踪。实验表明,可以实时、准确、稳定地对目标物体进行轮廓跟踪。
Real -time contour tracking algorithm can provide the contour of the object for video surveillance to classify the object or the action of the object. A new contour - tracking algorithm based on mean - shift and edge detection is proposed in this paper. The center of the object is got by using mean - shift tracking algorithm firstly, and at the same time the background is updated using Gauss model. Then the foreground rectangle and the background rectangle are obtained from the foreground and background image respectively. The foreground rectangle minus the background rectangle is the region of the object and then the contour of the object is got from the region of the object using edge detection method. Experiments show that this method can track the object timely, accurately and stably.
出处
《计算机仿真》
CSCD
2008年第6期224-227,共4页
Computer Simulation
关键词
均值漂移
边缘检测
轮廓跟踪
Mean shift
Edge detection
Contour tracking
