一种新的运动目标快速检测分割方法的研究

ON A NOVEL ALGORITHM OF FAST DETECTING AND SEGMENTING MOVING OBJECT

运动目标的检测与分割是视频分析的重要内容。对静态背景中的运动对象的检测方法进行了研究,针对基于混合高斯模型的背景减除法无法解决的"鬼影"和算法复杂耗时的问题,提出了一种新的基于帧差运动边缘检测的方法。实验证明,该方法可以在复杂背景下准确地获得运动对象边界,大大提高检测速度,同时有效消除背景光照变化及个别景物扰动带来的干扰。采用模板填充算法分割运动目标,并通过数学形态学滤波去除运动区域内的噪声点和填补空洞,获得完整理想的运动对象区域。

Detecting and segmenting moving object is an important subject in computer visual analysis.In this paper we studied the algorithms of detecting moving target in static background.As the existing background subtraction method based on mixture Gaussian model can＇t solve the problems such as the ＂ghost＂ and complicated time consuming of the algorithm,a new method of moving edge detection is proposed based on differential between the adjacent frames.Experimental results show that the proposed algorithm can accurately get the edge of moving object in complex background and greatly expedites detecting speed.At the same time it eliminates efficiently the background illumination variation and the disturbance inflicted by the perturbation of individual scene.We used template filling algorithm to segment the moving object,and used mathematic morphological filtering to remove noise points within the moving region and to fill up cavity,a complete and ideal moving object region is achieved.