摘要
为了消除静止摄像机下复杂背景环境对运动目标检测造成的影响,同时又为解决目前的运动目标检测算法在检测速度、准确度方面存在的不足,提出一种改进码本模型的运动目标实时检测算法.首先利用分块的思想将一幅图像分成若干宏块(MacroBlock),然后对每个宏块进行背景建模聚类成码本.在目标检测时,同样对当前帧进行分块,对于每个宏块求得像素均值和亮度均值并与其对应位置的码本进行比较判断,从而提取出前景目标.实验证实,对存在动态因素的背景视频,该算法不仅能有效抑制伪目标的出现,而且能准确快速地检测出运动目标.
In order to solve the problem on detecting moving objects with the complex background captured using a static camera,and at the same time,to overcome the shortage of detection speed and the accuracy in the detection of moving objects currently,we presented a real-time algorithm for the moving object detection with an improved codebook model.First,we divided an image into several blocks(MacroBlock) by using the idea of block.Then,for each macroblock,a codebook was built.During the moving object detection,we used the same way to divide the current frame into several blocks,just achieving each macroblock's pixel-means and brightness-means to compare with the codebook to identify whether it was foreground block or not.The experiment proves that this algorithm not only can restrain the fake target effectively but can also achieve robust and fast detection in videos which contain moving background.
出处
《中国计量学院学报》
2012年第2期125-130,共6页
Journal of China Jiliang University
基金
浙江省科技计划项目(No.2008C21135)
关键词
背景建模
运动目标检测
码本模型
宏块
background modeling
moving object detection
codebook model
macroblock
