一种新型的实效运动目标检测方法  

Novel and Efficient Moving Target Detection Method

作　　者：王红茹[1] 季鸣[1] 

机构地区：[1]江苏科技大学机械工程学院,镇江212003

出　　处：《计算机系统应用》2015年第12期208-214,共7页Computer Systems & Applications

基　　金：江苏科技大学高级人才项目(35020902)

摘　　要：提出一种新型的帧间差分光流的运动目标检测方法.该方法通过改进七帧差分和改进背景减除消除运动目标检测时出现的"空洞"和虚假目标;通过在光流计算方程加入权函数和引入通用动态图像模型建立新的光流约束条件,以解决常用光流场计算耗时长和亮度变化引起的约束方程不成立的问题,同时获取运动准确信息;最后通过阈值分割和形态学处理完成对目标的分割.实验对比分析表明,该方法能实现运动目标的准确快速检测与分割.A novel and efficient moving target detection method is proposed. The method by improving the seven-frame difference and background subtraction to eliminate the moving object detection of ＂empty＂ and false targets. Through the optical flow computation model for the weight function and the introduction of universal dynamic image model to establish a new optical flow constraints, it can solve common optical flow field calculation time-consuming and the unestablished problem caused by brightness changes of constraint equation, and also obtain accurate information campaign; Finally, threshold segmentation and morphological processing is completed to extract the target. Analysis indicate that the proposed method can detect and segment the moving target accurately and fastly.

关 键 词：运动目标检测 七帧差分 背景减除 光流场 通用动态图像模型 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]