改进运动模板与直方图匹配的视频典型目标检测方法  

Typical Video Target Detection Method Based on Improved Motion Template and Histogram Matching

作　　者：张永梅 冯超 马健喆 

机构地区：[1]北方工业大学计算机学院,北京 [2]北京华龙通科技有限公司,北京

出　　处：《计算机科学与应用》2019年第1期106-118,共13页Computer Science and Application

基　　金：国家自然科学基金(61371143);北方工业大学2018年教育教学改革和课程建设研究项目(18XN009-002);教育部高等教育司产学合作协同育人项目(201801121002);北方工业大学优势学科项目(217051360018XN044).

摘　　要：针对经典运动模板法提取运动目标轮廓出现的空洞和不能提取完整轮廓的问题,以及运动速度缓慢时误检率高的问题,本文提出了一种改进的运动模板和HSV直方图匹配相融合的视频运动目标检测方法。采用三帧差分和改进的OSTU自适应阈值法获得潜在的运动目标轮廓,并根据每个像素点与目标轮廓滞留时间的时序关系构建运动历史模板,利用逐级洪水泛滥法分割目标潜在区域,并计算目标潜在区域与目标模板的Bhattacharyya距离,从而检测典型目标区域。实验结果表明,该方法有助于解决传统运动模板法提取运动目标轮廓产生的空洞和不能提取完整轮廓的问题,而且精确率比背景减除法、运动模板法明显提高。The existing video moving target detection algorithms cannot detect the complete regions of moving targets and the hole phenomena in extracting the contour of video moving targets using classical motion template method and the false detection rate is higher when slower moving speed, the paper presents a video moving target detection method based on fusion of improved motion template and HSV histogram matching, adopts the three-frame difference and improved OSTU adaptive threshold method to obtain the contours of the potential moving targets, constructs the motion history template according to the time series relationship between each pixel point and the target contour retention time, segments the potential moving target areas using the stepwise flooding method, and calculates the Bhattacharyya distance between the potential target area and the target template area so as to detect the typical targets. The experiment results show the proposed method can help to solve the problems of the hole phenomena and incomplete contours for the typical target using classical motion template, and the average rate of detection accuracy is obviously higher than that of the background subtraction method and the motion template method respectively.

关 键 词：改进的运动模板 三帧差分法 改进的OSTU自适应阈值 逐级洪水泛滥法 BHATTACHARYYA距离 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]