复杂背景下运动目标分割算法研究  

Segmentation of Moving Targets in Complex Background

作　　者：万海东[1] 张骏[1] 刘真峥[1] 

机构地区：[1]西北工业大学,陕西西安710072

出　　处：《计算机仿真》2008年第11期282-284,324,共4页Computer Simulation

摘　　要：视频监控中,准确而快速地分割出运动物体是提取各种信息的前提和关键。以图像差分法为基础,提出了一种在复杂背景下有效分割运动目标的方法。运用改进的Surendra算法提取背景。对连续帧图像的R,G,B三通道帧差分图像采用PCA融合和二值形态学重构,以提取和更新背景。将粗分割图像转换到HSV域中,采用V分量阈值法消除阴影,并应用彩色投影法解决连通体粘连和路面反光问题。实验结果表明该方法能够结合各种算法的优势,快速而较准确地提取出运动车辆目标。Segmenting the targets correctly and quickly in video surveillance is the precondition for extracting all kinds of information. An algorithm of segmenting moving targets in complex background based on difference methods is proposed in this paper. The background is extracted by modified Surendra algorithm. The differences of two consecutive frames in channel R, G and B, converted into binary images, are fused by Principal Component Analysis method and morphological processing is applied to them to extract and update the background. The pre - segmented image is converted into HSV domain and the component V is weighted to eliminate the influence of shadow. Then the Color Projection algorithm is applied to disconnect the connecting target and suppress the glistening by the road. Experimental results show that the proposed algorithm can make use of different advantages to recognize the moving targets effectively and correctly.

关 键 词：背景提取 运动目标分割 阴影消除 彩色投影 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]