一种有效的运动物体检测方法

为了从监控视频中检测出较高质量的运动物体,文章提出了一种基于帧间差分和背景差分相结合的运动目标的检测方法,并且采用像素级和帧级背景更新相配合的一种背景更新策略。算法求取各像素点处的最大概率灰度,从而提取出连续视频的背景图像;相邻帧则利用帧间差分法以及背景差分法得到两幅运动区域图像;将两幅运动区域图像相与,提取出较为准确的运动目标。实验证明,该算法对光线的变化鲁棒性较高,运算速度较快,且能够及时的响应监控视频的实时变化,提高运动目标的检测质量。

A Effective Algorithm of Moving Object Detection

To extract good moving object from surveillance video, this paper proposes an algorithm of background extraction based on both flame-difference and background subtraction, and using multi-level adaptive background update strategy. Reconstructed the background image of continuous video frequency by calculating the maximum probability grayscale. Gained two moving regions by frame difference and background subtraction, took out moving objects by logic and the two subtracting moving regions. Experiment shows it is a practical method that is robust under lighting variations and computing speed. The algorithm can response timely to the surveillance video changes and improve the quality of moving objects detection.