基于改进的YOLO V3算法汽车零件配置辨识  

Identification of Vehicle Parts Configuration Based on Improved YOLO V3 Algorithm

作　　者：张丽秀[1] 田甜 邵萌[1] ZHANG Li-xiu;TIAN Tian;SHAO Meng(School of Mechanical Engineering,Shenyang Jianzhu University,Shenyang 110168,China;National-Local Joint Engineering Laboratory of NC Machining Equipment and Technology of High-Grade Stone,Shenyang Jianzhu University,Shenyang 110168,China)

机构地区：[1]沈阳建筑大学机械工程学院,沈阳110168 [2]沈阳建筑大学高档石材数控加工装备与技术国家地方联合工程实验室,沈阳110168

出　　处：《组合机床与自动化加工技术》2020年第6期150-153,共4页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：辽宁省重点研发计划项目(2017225016);沈阳市自然科学基金“建筑工程及材料分析测试服务平台建设”(Z3316006)。

摘　　要：为了提高检测的准确率并缩短检测时间以提高时效性,提出了一种以基于回归的目标识别方法YOLO V3算法为基础,将汽车前脸图像中的格栅、雾灯以及轮毂为目标,对YOLO V3网络结构参数进行优化,改进成为BBO-YOLO V3算法。对装配车间尾线工位进行了研究,检验在不同工况下,汽车前脸图像的识别效果。实验结果表明,该方法可成功代替人工检测,在生产节拍内实现检验,满足生产要求,从而能够提升生产车辆的质量,提高客户满意度。The final line station of the assembly workshop needs to make a judgement on whether the front face parts is correct or not,which is of great signification to reduce the working intensity and reduce the error inspection and missed inspection rate.In order to improve the accuracy of detection and shorten the detection time to improve the effectiveness,this paper based on the regression target recognition method YOLO V3 algorithm,took the grille,fog lights and wheel hubs of the car′s front face image as targets,optimized the network structure parameters of YOLO V3,and improved it to BBO-YOLO V3 algorithm.In this paper,the final line of the assembly workshop of BMW is taken as an example to test the recognition effect of front face image under different conditions.Experimental results show that this method can replace manual detection,and achieve detection within the production,Beat to meet production requirements,so as to improve the quality of production vehicles and improve customer satisfaction.

关 键 词：YOLO V3算法 零件识别 汽车质量控制 

分 类 号：TH162[机械工程—机械制造及自动化] TG506[金属学及工艺—金属切削加工及机床]