基于机器视觉的焊缝缺陷检测及分类系统的研究

发布时间：2018-05-08 01:02

本文选题：机器视觉 + 焊缝缺陷检测　；参考：《江南大学》2017年硕士论文

【摘要】：焊接技术是一门非常重要的基础工艺学科,它在现代制造业中起着重要作用。如今,无损检测手段如X射线成像等技术,已广泛应用于焊接工件的质量检测中。本文主要针对金属包装行业中常见的薄壁金属罐的焊缝检测技术进行研究。由于薄壁金属罐的特性,常用的无损检测手段并不适用,而人工抽样检测是目前企业在实际生产中最为常用的检测手段。但这种方式比较依赖于检测人员的经验,且效率低、具有随机性,极大的影响了企业生产的效率及产品的质量。受到人工目测检测方式的启发,本文对视觉检测的方式进行了大量研究,设计了基于机器视觉方式的焊缝缺陷检测及分类系统。系统通过工业摄像机等设备采集焊缝样本图像,通过图像预处理操作对样本进行初步处理,然后使用改进的背景差分和波形检测等方法对焊缝缺陷进行了检测及类型判别。具体工作如下:首先,本文对系统的整体框架进行了介绍,重点介绍了图像采集系统,并简单介绍了焊缝图像处理的软件流程。此外,还对熔焊、虚焊、焊穿这三类典型的焊缝缺陷的成因及特点进行了介绍。其次,本文介绍了焊缝图像预处理的流程。预处理流程主要包括对图像的筛选、对图像有效区域的提取、图像旋转矫正以及焊缝核心区域的提取。图像预处理能够排除大部分离散伪缺陷的干扰,且将罐身区域焊缝图像与两端焊缝图像区分开来,提取了图像的核心区域。然后,本文详细阐述了罐身区域焊缝缺陷的检测与类型判别流程。本文提出了改进的背景差分法,用于构建焊缝图像序列的背景模型,提取焊缝图像的缺陷特征,并通过这些特征在面积、亮度、累加波形等方面的区别,对缺陷类型判别设计了相应算法。最后,本文阐述了两端焊缝缺陷的检测方法。在这一章中提出了均值阈值分割的方法消除了残影干扰,使用拟合波形方法判断焊缝缺陷,并通过横向累加波形的方法检测虚焊缺陷。在线检测实验结果表明,本文设计的整个焊缝缺陷检测与分类系统达到了99%以上的准确率,能够满足企业生产的实际需求。
[Abstract]:Welding technology is a very important basic technology subject, it plays an important role in modern manufacturing. Nowadays, nondestructive testing techniques, such as X-ray imaging, have been widely used in quality testing of welded workpieces. This paper mainly focuses on the weld inspection technology of thin-wall metal tank in metal packaging industry. Because of the characteristics of thin-walled metal tank, the commonly used nondestructive testing method is not suitable, and manual sampling testing is the most commonly used testing method in the actual production of enterprises at present. But this way depends on the experience of the examiner, and the efficiency is low, which has randomness, which greatly affects the production efficiency and the quality of the product. Inspired by the manual visual inspection method, the visual inspection method is studied in this paper, and a weld defect detection and classification system based on machine vision is designed. The system collects the weld seam sample image through the equipment such as the industrial camera, carries on the preliminary processing through the image preprocessing operation, and then uses the improved background difference and the waveform detection method to detect the weld seam defect and to distinguish the type. The main work is as follows: firstly, the whole frame of the system is introduced, the image acquisition system is introduced, and the software flow of weld image processing is briefly introduced. In addition, the causes and characteristics of three kinds of typical weld defects, such as weld welding, virtual welding and weld penetration, are also introduced. Secondly, this paper introduces the process of weld image preprocessing. The process of preprocessing mainly includes image screening, extraction of image effective region, image rotation correction and extraction of weld core area. The image preprocessing can eliminate the interference of most discrete pseudo-defects and distinguish the weld image of the tank body region from the weld image of both ends and extract the core area of the image. Then, this paper expatiates the process of weld defect detection and type discrimination in the tank body area. In this paper, an improved background difference method is proposed to construct the background model of the weld image sequence, extract the defect features of the weld image, and through the differences of these features in area, brightness, cumulative waveform, etc. The corresponding algorithm is designed for the discrimination of defect types. Finally, this paper describes the detection method of weld defects at both ends. In this chapter, the mean value threshold segmentation method is proposed to eliminate the residual interference, and the fitting waveform method is used to judge the weld defect, and the transverse accumulative waveform is used to detect the virtual welding defect. The on-line testing results show that the whole weld defect detection and classification system designed in this paper has the accuracy of more than 99% and can meet the actual requirements of the enterprise production.
【学位授予单位】：江南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG441.7;TP391.41

【参考文献】