【作者】 李明；

【导师】 毛建旭； 段峰；

【作者基本信息】 湖南大学 ， 电子与通信工程（专业学位）， 2017， 硕士

【摘要】 随着“中国制造2025”的概念深入人心,智能制造的理念将成为未来汽车工业制造的主流方向。机器视觉不仅能够增强工业机器人的自主能力和灵活性,而且能够使自动化产线转型升级更趋于智能化。汽车挡风玻璃定位与支架粘合作为汽车整车制造的一个重要环节,其安装精度与质量是决定整个汽车安全性的重要性能指标之一。目前,国内各企业对汽车挡风玻璃定位与支架粘合检测主要采用人工或者机械的方法,存在加工时间长、效率和粘合精度低、人工成本高、制造加工人员劳动强度大且容易疲劳等弊端,严重影响生产企业的效率,迫切需要研制基于机器视觉的汽车挡风玻璃定位与支架粘合智能检测系统。本文首先介绍了课题研究工作的背景和意义,对目前国内外机器视觉技术的现状、特点以及应用领域进行了阐述。然后根据汽车生产厂家对视觉检测系统的需求,提出了汽车挡风玻璃定位与支架粘合视觉检测系统的整体设计方案,详细介绍了电气控制系统、机器人控制系统、视觉检测系统的设计与实现。接着,针对汽车挡风玻璃视觉定位涂胶的要求,研究并实现了基于自适应阈值分割和改进霍夫变换的挡风玻璃视觉定位算法,采用局部自适应分割算法对图像数据进行分割,采用改进的霍夫变换对梯形与圆形特征进行特征信息提取。实验证明,自适应阈值分割算法和改进的霍夫变换方法能够有效高精度的提取挡风玻璃特征信息,其定位精度小于厂家要求的0.2mm。然后,设计了汽车挡风玻璃支架粘合精度检测算法。设计过程中先利用像素累加算法确定支架检测区域坐标系以及检测区域的定位,再利用Blob分析工具得到面积占比与缺陷内接矩形的长度和宽度等信息判定支架底漆与涂胶的质量,最后通过基于梯度值与改进的霍夫变换提取粘合后支架圆心位置以及几何检测算法,验证支架粘合的精度是否符合厂家精度要求。实验结果表明,该系统在底漆与涂胶的误检率和漏检率上都比人工检测要更优良。在加工时间上,人工检测是智能系统的4.6倍;人工检测精度只有77.5%,远低于智能系统的95.5%。最后,开发了视觉检测系统软件,详细阐述了视觉检测软件的流程和各个软件子功能模块的设计与实现,介绍了子模块函数库的调用以及软件通信方式,实现了汽车挡风玻璃视觉定位涂胶和支架粘合质量在线检测。该视觉检测软件能够高效的应用于工业现场,达到高速、高精度的生产需求。更多还原

【Abstract】 With the concept of "Made in China 2025",the concept of intelligent manufacturing will become the mainstream of the future automobile industry manufacturing.Machine vision can not only enhance the autonomy and flexibility of industrial robots,but also to make automation line transformation and upgrading more intelligent.Automotive windshield positioning and bracket bonding as an important part of the vehicle manufacturing,the installation accuracy and quality of the car is to determine the safety of one of the important performance indicators.At present,the domestic enterprises on the car windshield positioning and stent adhesion testing mainly using artificial or mechanical methods,the existence of a long processing time,low efficiency and bonding accuracy,high labor costs,manufacturing and processing personnel labor intensity and easy fatigue defects,seriously affect the efficiency of production enterprises,the urgent need to develop based on the machine vision of the car windshield positioning and stent adhesion intelligent detection system.Firstly,This paper introduces the background and significance of the research work,and expounds the present situation,characteristics and application fields of machine vision technology both at home and abroad.Then,according to the requirements of the visual inspection system of the automobile manufacturer,the overall design scheme of the vehicle windshield positioning and the bracket visual inspection system is put forward.The design and implementation of the electrical control system,the robot control system and the visual inspection system are introduced in detail.Then,aiming at the requirement of visual positioning glue coating for automobile windshield,the adaptive localization algorithm of windshield based on adaptive threshold segmentation and improved Hough transform is studied and realized.The local adaptive segmentation algorithm is used to segment the image data,The Hough transform is used to extract the characteristic information of trapezoidal and circular features.Experiments show that the adaptive threshold segmentation algorithm and the improved Hough transform method can effectively and accurately extract the characteristic information of the windshield,and the positioning accuracy is less than 0.2mm of the manufacturer’s requirement.Then,the design of the comfort of the windshield bracket is designed.In the design process,the pixel accumulation algorithm is used to determine the position of the bracket detection area coordinate system and the detection area.Then,the Blob analysis tool is used to determine the area and the length and width of the insured rectangle.Finally,the accuracy of the bonding of the stent is verified by the gradient value and the modified Hough transform,and the geometric detection algorithm is used to verify the accuracy of the bonding.The experimental results show that the system is better than the manual detection of the false detection rate and the missed detection rate of the primer and the coating.In the processing time,the manual detection is 4.6 times the intelligent system;manual detection accuracy of only 77.5%,much lower than the intelligent system 95.5%.Finally,the visual inspection system software is developed,the process of visual inspection software and the design and realization of each software sub-function module are described in detail.This paper introduces the call of sub-module library and the way of software communication,and realizes the on-line detection of car windshield visual positioning glue and stent bonding quality.The visual inspection software can be used effectively in the industrial field,to achieve high-speed,high-precision production needs.更多还原