基于光流与Mean Shift算法的运动目标检测

Detection of moving objects based on optical flow vector and mean shift algorithm

为了改善传统光流法不能很好地解决背景运动带来影响的问题,提出了一种基于光流矢量与Mean Shift算法的运动目标检测方法。结合图像中边缘信息和灰度信息,用金字塔法计算出特定像素点的Lucas-Kanade光流矢量。再利用Mean Shift算法的梯度搜索原理找出运动背景的光流矢量,进而找出运动目标的光流矢量。实验结果表明,融入Mean Shift算法后的光流法高精度地检测出了目标,显著降低背景运动所带来的影响。文中所提方法比传统几种方法更加准确地得到了所需结果。

In order to improve the traditional optical flow method which can't solve the problem of the impact of background movement,a new method of moving objects detection based on optical flow vector and Mean Shift algorithm is proposed. Using the edge information and gray information of the image,the Lucas-Kanade optical flow vector of a specific pixel is calculated by the Pyramid method. The optical flow vector of the moving background is found by using the gradient search principle of Mean Shift algorithm,and then it finds the optical flow vector of the moving target. The experimental results show that the optical flow method of Mean Shift algorithm finds out the moving vector of the objects with high precision and obviously reduces the impact of background movement. In this paper,the proposed method is more accurate than the traditional methods.