基于连通性检测的视频监控运动目标提取

针对背景相对静止、主要检测对象为行人的视频监控序列,提出了一种基于连通性检测的目标提取算法,它把形态学滤波与连通性检测相结合,对分割后的二值化图像进行噪声干扰去除,在获得若干连通区域后,利用面积、外界矩形及其特征对连通区域进行识别,通过区域重心标注目标在各帧位置,实现目标提取.实验结果表明,该算法简单可靠,具有实时性,易于硬件实现,可应用于实际系统.

Moving Target Extraction of Video Supervision Based on Connectivity Detection

According to the video supervision sequence, in which the background is relatively still and the objects to be detected are pedestrians ,an algorithm of moving target extraction based on connectivity detection is proposed. By this method, which combines morphology filtering with connectivity detection, the noise and interference of the binary image can be removed. The object regions are expressed with region area,and according to the bounding rectangle features of pedestrians,the algorithm can identify the target from several object regions and calibrate the target location by gravity coordinate. Experimental result shows that this algorithm can extract targets from video supervision sequence effectively and simply and can be applied to real-time hardware system.