Sobel算子在提取视频运动目标轮廓中的应用研究

在计算机视觉领域中,从视频流中获取清晰的运动目标轮廓对于目标的跟踪识别以及行为判断等高层次处理具有非常重要的意义.传统的帧间差分法虽然可以粗略的提取到运动目标,但却未能很好的解决算法留下的“空洞”和“重影”问题.本文首先采用Sobel算子提取每一幅帧图像的轮廓,然后将帧间差分法提取得到的运动目标同该轮廓做“与”运算,从而得到线条清晰的目标轮廓.实验表明,该方法较传统帧间差分法而言,在保证原有轮廓的同时,也能消除移动速度对轮廓的影响.

A research of extracting the contour of moving target using Sobel operator

In the digital video monitoring system,To capture a clear contour of moving target from the video stream is important and significative for high-level processing like target tracking and recognition of behavior.Traditional method of interframe difference may extract the moving target roughly,but does not do better int solving the ＂cavity＂ and ＂double＂.this paper firstly extracts the contour of every frame image using Sobel operator,then extacts a more clear contour with ＂and＂ operation by using interframe difference and the result above,finally gets a mor stable and clear contour of a moving target.Experiments show that this method can not only extract the contour effectively,but also eliminate the influence of moving target＇s speed.