基于区域生长处理的视频运动车辆提取新方法  

New method of video-based moving vehicle extraction based on seeded region growing

作　　者：何得平[1] 赵广州[1] 陆承恩[1] 朱光喜[1] 

机构地区：[1]华中科技大学武汉光电国家实验室,武汉430074

出　　处：《计算机应用》2008年第4期979-981,共3页journal of Computer Applications

摘　　要：车辆提取在视频交通流量检测系统中是最重要、最关键的一环。提出一种基于区域生长算法的简单、快速的车辆提取方法。这种方法通过差分图像估计阴影灰度近似值,然后用该值作为参考值删除输入图像属于背景和阴影的像素,最后得到运动车辆图像。利用该方法进行了大量的实验,从实验结果来看,此方法具有较好的稳定性和有效性;从计算时间来看,该方法也能满足实时系统的需要。It is an important and key part to extract moving vehicle in intelligence transportation system. A new, simple and fast method was proposed to extract moving vehicle from video based on seeded region growing. Gray approximation was estimated from difference image, and all pixels belonging to background and shadow were removed. Experiment results confirm that this proposed method is efficient and reliable, and its computation cost can also satisfy the vehicle recognition system.

关 键 词：背景 阴影 运动车辆 灰度近似值 边缘点 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]