汽车前底盘装配视觉检测系统设计与应用  

Design and Application of Vehicle Front Chassis Assembly Visual Inspection System

作　　者：李硕 苑明哲[2,3] 王文洪[3] 史洪岩 肖金超 宋纯贺 曹飞道[3] LI Shuo;YUAN Mingzhe;WANG Wenhong;SHI Hongyan;XIAO Jinchao;SONG Chunhe;CAO Feidao(College of Information Engineering,Shenyang University of Chemical Technology;Shenyang Institute of Automation,Chinese Academy of Sciences;Guangzhou Industrial Intelligence Research Institute)

机构地区：[1]沈阳化工大学信息工程学院 [2]中国科学院沈阳自动化研究所 [3]广州工业智能研究院

出　　处：《仪表技术与传感器》2024年第6期79-85,共7页Instrument Technique and Sensor

基　　金：国家自然科学基金面上项目(62273332,62273337);中科院科技服务网络计划(STS)-东莞专项(20211600200072)。

摘　　要：为解决汽车底盘混流装配错装、漏装和人工检测效率低的问题,设计了基于YOLOv3-Tiny的在线检测系统。该检测系统利用4套光源-相机组合的成像系统,从多角度获取前底盘模块的全貌图像,利用基于差分统计的条纹识别算法剔除低质量图像;根据检测目标特性,简化非极大值抑制算法,优化检测过程。实验和现场运行结果表明:检测系统目标无遮挡检出率达到100%,综合识别准确率达到99.95%,平均检测时间3.5 s,较之前人工检测效率提升94.55%,检测系统具有较高的准确度和检测效率,在汽车工业中实现了柔性化和智能化的目标检测应用。In order to solve the problems of wrong assembly,missing assembly and low efficiency of manual inspection in au-tomobile chassis mixed-flow assembly,an on-line detection system based on YOLOv3-Tiny was designed.The detection system used four sets of imaging systems consisting of light sources and cameras to obtain the panoramic image of the front chassis mod-ule from multiple angles.And the detection system used the fringe recognition algorithm based on differential statistics to eliminate the low-quality image.According to the characteristics of the detection target,the non-maximum suppression algorithm was simpli-fied.And the detection process was optimized.The results of experiment and field operation show that the unoccluded detection rate,comprehensive recognition accuracy and average detection time are 100%,99.95% and 3.5 s.The average detection time is 94.55%lower than that of manual detection.The detection system has higher accuracy and higher efficiency,and has achieved flexible and intelligent target detection applications in the automotive industry.

关 键 词：汽车制造 汽车装配部件检测 YOLOv3 条纹检测 非极大值抑制 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]