基于差分法和概率估计方法的运动目标检测  

Moving objects detection based on the method of image difference and its probability estimation in histogram

作　　者：苑泊舟[1] 孙杰[1,2] 程海林[1] 马冉冉[1] 贺芳[1] 马斌[2] 

机构地区：[1]天津理工大学计算机与通信工程学院天津市薄膜电子与通信器件重点实验室,天津300384 [2]天津理工大学电子信息工程学院,天津300384

出　　处：《天津理工大学学报》2012年第1期63-67,共5页Journal of Tianjin University of Technology

基　　金：天津市自然科学基金(06YFJMC15600)

摘　　要：针对视频序列运动目标检测问题,提出根据灰度直方图概率密度函数特点自动确定阈值的方法,以准确地从背景中分割出运动目标.本文以差分法为基础,利用正态分布的经验法则确定阈值,对差分图像进行分割,并用方框标识运动目标.该方法省去了滤波、形态学处理等冗余步骤,缩短了图像处理的时间.实验表明,该方法实时性强,且对面积较大或较小的运动目标都能达到有效跟踪的目的.A new algorithm,the threshold was chosen automatically according to the characteristics of the distribution probability in the grayscale histogram,was given for the segmenting processing of the moving object from background in solving the problem of the moving target detection in video sequence.Based on the image difference,the outline of the moving target was gotten by the segmentation with the threshold fixed automatically according to the experiment rule of normal distribution.A box is used to mark the moving objects location.The filtering,morphological processing that waste a lot of time is not needed while this method is used and it effectively short the time of the image processing for target detection.Experimental results indicate that this algorithm is good at real-time detection of moving object and there is no limitation on size of the moving target.

关 键 词：图像处理 运动目标检测 自动阈值选择 对数正态分布 差分法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]