一种基于帧差分法与快速图分割相结合的运动目标检测方法  

A moving target detection method based on integration of frame difference method and fast image segmentation

作　　者：陈佳[1,2] 朱长仁[1] 钱智明[1] 

机构地区：[1]国防科学技术大学电子科学与工程学院ATR重点实验室,湖南长沙410073 [2]武警福建总队直属支队,福建福州350001

出　　处：《现代电子技术》2016年第3期13-17,22,共6页Modern Electronics Technique

基　　金：国防科大ATR重点实验室项目资助;国防科大创新论文资助

摘　　要：针对在处理复杂场景情况时,帧差分法在预处理后获取的目标区域往往会空洞化或者分裂成多个互不相连的部分,计算机难以实时判断多个部分是否同属于一个目标,为此,提出一种基于三帧差分法与快速图分割相结合的运动目标检测方法。首先对序列图像做三帧差分运算及基于图的快速分割;然后通过定义融合判定规则,将帧差分结果同图像分割的区域块进行互相补充和约束;再对符合判定规则的子区域块进行合并,最终精确提取出运动目标的位置及轮廓。实验结果表明该方法能够有效提高目标检测的完整性。The target area acquired by the frame difference method after pretreatment becomes hollowing or is split into several disconnected parts,and it is difficult for computer to judge whether the multiple parts belong to one object,so a moving target detection method based on integration of three-frame difference method and fast image segmentation is proposed. The three-frame difference operation and image-based fast segmentation for sequential images are conducted,and then the decision rule is fused by definition to carry out mutual supplementation and restraint of the difference results and image segmentation region block,after that,the sub-region blocks corresponding to the decision rule are combined,and the position and outline of the moving target are extracted precisely. The experimental results show this method can effectively improve the integrity of the moving target detection.

关 键 词：目标检测 帧差分 图像分割 最小生成树 视频监控 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]