基于彩色图像边缘差分的运动目标检测算法

Moving object detection algorithm based on color image edge-differencing extraction

针对目前常用的运动目标提取易受到噪声影响、易出现阴影和误检漏检等情况,提出了一种基于Sobel算子的彩色边缘图像检测和帧差分相结合的检测方法。首先用Sobel算子提取视频流中连续4帧图像的彩色边缘图像,然后将边缘图像进行隔帧差分相与,提取出较精确的运动目标边缘轮廓。提取的轮廓经过一系列的形态学操作填充,可得到完整的运动目标。实验结果表明,该方法对运动目标边缘轮廓提取准确,抗噪能力强,且鲁棒性好。

To improve the moving object detection which susceptible to noises and shadows, which has incorrect leak detection, an algorithm based on color edge extraction of Sobel, kenal and frame differencing is presented in this paper. Firstly, four continuous frames of color edge image are extracted by the edge extraction from video. Then, the edge-image-differencing is made so as to extract edge contours of moving object. Finally, the edge contours are filled with morphology method. The test results shows that the proposed algorithm can detect the moving object accurately and quickly with less noise.