交通流视频检测中背景模型与阴影检测算法  

Background extraction model and shadow detection algorithm in traffic flow video detection

作　　者：李志慧[1] 张长海[2] 曲昭伟[1] 王殿海[1] 

机构地区：[1]吉林大学交通学院,长春130022 [2]吉林大学计算机科学与技术学院,长春130012

出　　处：《吉林大学学报（工学版）》2006年第6期993-997,共5页Journal of Jilin University:Engineering and Technology Edition

基　　金：国家自然科学基金资助项目(50422285);吉林省科技厅国际合作处资助项目(20040705-2);国家人事部归国优秀人员基金资助项目

摘　　要：提出了基于对象级的混合高斯背景模型更新方法与基于RGB颜色变化度的运动阴影检测算法。根据运动分割、物体识别、Kalman运动跟踪等高层语义表达,结合像素的时空特性,进行基于对象级的混合高斯背景更新。克服了像素级混合高斯模型中交通控制信号或交通阻塞等造成的长时间停车以及交通高峰期交通拥挤等情况下对背景抽取造成的影响;根据运动目标的RGB颜色变化度特点,提出自适应的对象级运动阴影检测算法,克服了运动阴影的影响及其造成的误分类。不同交通状态下的视频处理效果表明,该方法具有良好的鲁捧性和自适应性。To alleviate the difficulties in the detection and recognition of the moving objects, even the possibility of the object misclassification, due to the effect of the variation of the moving object shadow and the background factors, a mixed Gaussian background update model based on the object level and a moving object shadow detection algorithm based on the RGB color variation degree were proposed. It performs the mixed Gaussian background update according to the object high-level semantic expressions, such as movement segmentation, object recognition, Kalman movement tracking, etc. , in the light of spatio temporal features of the pixels, eliminates the effect of the prolonged traffic standstill due to the traffic control signs and the rushtime traffic congestion on the background extraction in the mixed Gaussian model based on the pixel level, avoids the object misclassification due to effect of the moving shadows. The results of experiment on the video pictures of different traffic conditions showed the proposed technique is robust and self-adaptive.

关 键 词：计算机应用 视频检测 背景提取模型 阴影检测 交通流检测 

分 类 号：U121[交通运输工程] TP391[自动化与计算机技术—计算机应用技术]