改进五帧差法机器视觉气密性检测研究  

Airtightness Detection with Improving Inter Five Frame Difference Machine Vision

作　　者：刘纪飞 姜周曙[1] 陈磊 倪誉航 陈庆华 林剑 刘若清 LIU Jifei;JIANG Zhoushu;CHEN Lei;NI Yuhang;CHEN Qinghua;LIN Jian;LIU Ruoqing(School of Automation,Hangzhou Danzi University,Hangzhou 310018,China;Zhejiang Santian A/C Compressor Co.,Ltd,Longquan 323700,China)

机构地区：[1]杭州电子科技大学自动化学院,杭州310018 [2]浙江三田汽车空调压缩机有限公司,龙泉323700

出　　处：《真空科学与技术学报》2024年第5期445-451,共7页Chinese Journal of Vacuum Science and Technology

基　　金：浙江省“2024年度浙江省“尖兵”“领雁”研发攻关计划”项目(2024C01229)。

摘　　要：气密性是真空设备最重要的技术指标。在工业设备检漏中,传统皂泡法依赖人工目视,其检测准确性和效率取决于工作人员的专注度和操作技能。文章采用机器视觉技术,基于五帧差目标检测算法,进行检漏和漏点定位。由于传统帧差法存在部分画面重叠、边缘信息缺失和处理高帧率视频效果较差等问题,文章提出用Otsu算子和Canny边缘检测共同改进五帧差法。首先对连续图像做差分并转化为二值图,然后对中间帧进行Canny边缘检测,将二值图和Canny边缘检测的结果进行逻辑运算、中值滤波和形态学处理,最后获得完整的皂泡目标。实验表明,新算法能够处理高帧率视频并获得更完整的皂泡轮廓信息,可用于机器视觉真空设备气密性检测。Air tightness is the most important technical indicator of vacuum equipment in industrial equipment leak detection.The commonly used traditional soap bubble method relies on manual visual inspection,and its detection accuracy and efficiency depend on human focus and operational skills.Machine vision technology is applied in leak detection and leakage point localization based on the five-frame differencing method in this thesis.Due to the partial overlap and missing edge information using traditional frame difference methods,as well as poor performance in processing high frame rate videos,an improved inter five frame difference method using Otsu and Canny edge detection is proposed.Firstly,the continuous image is differentiated and converted into a binary image.Then,Canny edge detection is performed on the intermediate frames,and the binary image and edge detection results are subjected to logical operations,median filtering,and morphological processing.Finally,the complete soap bubble target is obtained.Through experiments,high frame rate videos can be processed,and more complete soap bubble contour information can be obtained in the new algorithm,which can be used for machine vision vacuum equipment airtightness detection.

关 键 词：气密性 五帧差法 边缘算法 运动目标检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]