基于视频的运动目标检测算法的比较与分析

Comparison and Analysis of Video-based Moving Object Detection Algorithms

帧差法和背景差分法是目前常用的基于视频的运动目标检测方法。首先从帧差法和背景差分法中选取了几种具有代表性的算法;然后针对典型场景条件对这些方法进行了实验测试,并利用实验结果对算法进行了定性和定量的比较和分析。实验结果表明,ViBe算法、码本模型、混合高斯模型、自适应混合高斯模型、中值滤波和自适应背景建模具有较好的检测效果,但任何一种方法都有其局限性。

Frame subtraction and background subtraction methods are widely used in video-based moving object detection methods. First, several typical algorithms are selected from frame subtraction and background subtraction methods. Then, experiments are conducted for several typical background conditions, and the experimental results are shown for qualitative and quantitative comparison and performance judgement. Experimental results indicate that the detection results of ViBe algorithm, codebook model, mixture Gaussian model, adaptive mixture Gaussian model, median filter and adaptive background model are better than others, but any method has its limitation.