融合五帧差分与高斯模型的运动物料袋检测  

Moving material bag detection method of a fused five frame difference and Gaussian model

作　　者：贾伟[1] 黄小天[1] 谢椿[1] 王正勇[1] 

机构地区：[1]四川大学电子信息学院图像信息研究所,四川成都610064

出　　处：《电子技术应用》2013年第10期139-142,共4页Application of Electronic Technique

摘　　要：提出了一种快速检测方法。以动态传送带与物料袋为研究对象,用改进的五帧差分法提取运动物料袋图像,同时以中间帧图像为参照进行单高斯背景更新;采用中间帧图像与背景图像做差,将得到的前景图像与帧差法得到的运动图像进行形态学"与"运算;当帧数满足给定条件时,表明该帧为完整物料袋图像,保存帧图像,并对不满足该条件的相邻帧图像通过加运算补偿。实验结果表明,该方法可以得到准确、快速的检测结果,并且对环境变化具有较强的鲁棒性。The article proposes a rapid detection method.Firstly,the article takes the dynamic conveyor belt and the plastic bags as the research object,uses the improved five-difference method to extract the images of moving material bags,At the same time,it takes the intermediate frame picture as a reference to update background ;Then using the difference between the middle frame image and the background image,and makes morphological and operation between the foreground image and the moving image obtained by frame difference method;When the number of frames meets special condition,it shows that the frame includes complete image information of material bag,and saves the frame image,continues to detect;Oppositly,then compensate the adjacent frames which is not satisfied the condition by add operation.the experimental results shows that the method is accurate and fast,and has a strong robustness to environmental change.

关 键 词：五帧差分法 单高斯模型 形态学运算 物料袋检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]