摘要
基于视频的目标检测中,现有流行的高斯混合模型GMM(Gaussian Mixture Model)目标检测的效果最好,但是其计算量很大,而简单的帧间差分方法和背景差分方法计算速度快,但是检测效果较差。提出在改进聚类方法基础上的基于自适应域值混合差分的目标检测方法,能够一方面具有很好的运动目标检测效果,另一方面具有很快的计算处理速度。同时该方法具有自适应能力,免除人工设置域值的麻烦,因而在实践中具有良好的实际使用价值。
In target detecting methods based on video processing technology, GMM ( Gaussian Mixture Model) is today' s popular way with highest detecting effect but is time-consuming in computation. Meanwhile those methods such as inter-frame difference and background difference are simple and quick in computation but with poorer detection qualityIn this paper,we proposed a target detection method based on adaptive threshold mixed difference with the improved OpenCV clustering algorithm. The method was proved to have perfect detection effect on mo- tion targets and fast computing speed. Besides, it was also adaptive in getting rid of threshold manual setting, thus had the practical applied value in engineering.
出处
《计算机应用与软件》
CSCD
2009年第10期94-97,共4页
Computer Applications and Software
基金
广东省科技计划项目(2006B11301001)
广东省国际科技合作计划项目(2007A050100026)
广东省工业科技攻关计划项目(2006B80407001)
关键词
目标检测
高斯混合模型
背景差分
混合差分
自适应域值
Target detection Gaussian mixture model Background difference Mixed difference Adaptive threshold