改进的基于区域的运动目标分割方法

针对视频监控系统,提出了一种改进的基于区域的运动目标分割方法。与传统方法相比,在运动检测阶段,结合时域差分和背景差分进行运动检测,并通过自适应方法进行背景更新;在差分图像二值化时,采用自适应阈值方法来代替传统的手工确定阈值法;对于区域分割,使用基于加权平方欧式距离的均值聚类算法代替传统的均值聚类算法。实验结果表明该改进方法比传统方法具有更好的实时性、鲁棒性和有效性。

The paper presents an improved moving object segmentation method based on region with focus on a video monitoring system. The motion  detection method combines time difference with background difference compared with the traditional method and background images are updated by  adaptive method. It uses an adaptive thresholding method to choose the thresholding value automatically, instead of determining the thresholding value  manually. It applies weighted K-means clustering for region segmentation, instead of traditional K-means clustering. The experiment results show this  method is real-time , robust and efficient.

中国科学院沈阳自动化所开放课题基金资助项目;;山东省自然科学基金资助项目(Y2002G04)