基于机器视觉的玻璃纤维布缺陷检测技术研究
发布时间:2018-03-27 20:18
本文选题:玻璃纤维布 切入点:机器视觉 出处:《郑州大学》2017年硕士论文
【摘要】:随着经济的发展,我国纺织业步入迅猛发展的阶段。然而国内大多数纺织品生产企业劳动密集程度较高,纺织品的缺陷检测仍然依靠人工检测,这种方式存在主观性强,精确度低,工作强度高等诸多弊端。机器视觉技术的日益成熟,使其在工业生产过程中的应用越来越为广泛,基于机器视觉技术的纺织品在线缺陷检测已然成为纺织品质量控制的重要发展方向。国外的织物在线检测技术起步相对较早,然而从外国引进织物在线缺陷检测设备价格昂贵,成本较高。国内的研究主要是针对某一种算法的研究且仅适用于一种织物缺陷检测,不能直接应用于玻璃纤维布的缺陷检测生产实际中。因此研究玻璃纤维布缺陷检测的关键技术,对于推动玻璃纤维织物自动化生产和布匹质量快速分级具有极其重要的意义。本文对玻璃纤维布缺陷检测系统的关键技术进行了深入系统地研究,主要内容包括:基于玻璃纤维布的纹理特性、检测要求和生产环境等设计了玻璃纤维布缺陷检测系统的总体方案,搭建了玻璃纤维布机器视觉检测平台,根据玻璃纤维布的缺陷特征,确定了背光照明的光源配置方案和基于GigE的多相机检测方案,获得了高对比度织物图像,降低了缺陷识别难度。针对被检玻璃纤维布布幅较宽和工业CCD视场小的问题,提出采用多相机同步采集图像然后对采集到的多幅图像进行拼接处理的实用性方案。分别基于模板匹配拼接方法和基于Harris特征点拼接方法研究了玻璃纤维布图像拼接技术,并从配准精度、拼接速度等方面对二者进行了对比分析。本文从实时性和可靠性出发,选择了基于模板匹配的拼接方法进行玻璃纤维布图像的拼接工作。为了解决玻璃纤维织物在线检测效率低、实时性差等问题,提出了一种基于Blob分析的织物缺陷检测方法。首先对织物图像采用均值滤波器进行平滑处理,以削弱噪声和织物纹理的干扰,然后采用迭代法寻找最佳阈值将图像分割为Blob和背景的像素集合,采用形态学处理调整分割后的Blob形状,最后对图像进行连通性分析和特征提取,通过对Blob区域进行最小外接矩形拟合得到缺陷特征的个数和尺寸等信息,实现了玻璃纤维布劈缝、跳花、破洞、污渍等常见缺陷的识别。实验结果表明,该方法计算简单,检测结果稳健可靠,实时性好,是一种有效的织物缺陷在线检测方法。在VS2010平台下基于C#、Halcon和SQL Sever数据库研制了玻璃纤维织布缺陷检测软件系统,系统包括图像采集模块、人机交互模块、图像处理模块和缺陷数据统计模块,实现了玻璃纤维布缺陷的检测和布匹分级。在实验平台上进行了调试和实验验证,结果表明本文的研究方法稳定可靠、实时性好,满足了预期的研发要求。
[Abstract]:With the development of economy, China's textile industry has stepped into a stage of rapid development. However, most domestic textile production enterprises have a relatively high labor intensity, and the testing of textile defects still depends on manual detection, which has strong subjectivity. With the development of machine vision technology, its application in industrial production is becoming more and more extensive. Textile on-line defect detection based on machine vision technology has become an important development direction of textile quality control. The domestic research is mainly for a certain algorithm and only suitable for one kind of fabric defect detection. It can not be directly used in the production of glass fiber cloth defect detection. Therefore, the key technology of glass fiber cloth defect detection is studied. It is of great significance to promote the automatic production of glass fiber fabric and the rapid grading of fabric quality. In this paper, the key technology of glass fiber fabric defect detection system is studied deeply and systematically. The main contents are as follows: based on the texture characteristics, detection requirements and production environment of glass fiber cloth, the overall scheme of glass fiber cloth defect detection system is designed, and the machine vision testing platform of glass fiber cloth is built. According to the defect characteristics of glass fiber cloth, the light source configuration scheme of backlight illumination and the multi-camera detection scheme based on GigE are determined, and the high contrast fabric image is obtained. The difficulty of defect identification is reduced. Aiming at the problems of wide fabric width and small field of view of industrial CCD, In this paper, a practical scheme is proposed to synchronize the image acquisition with multiple cameras, and then the practical scheme of image stitching is proposed, which is based on template matching method and Harris feature point stitching method, respectively, and the glass fiber cloth image mosaic technology is studied respectively, which is based on the template matching method and the Harris feature point stitching method, respectively. The comparison and analysis of registration accuracy and stitching speed are also given in this paper, which is based on real-time and reliability. In order to solve the problems of low efficiency and poor real-time of on-line detection of glass fiber fabric, the paper chooses the mosaic method based on template matching to join the image of glass fiber cloth. In this paper, a fabric defect detection method based on Blob analysis is proposed. Firstly, the mean value filter is used to smooth the fabric image to reduce the noise and fabric texture interference. Then iterative method is used to find the best threshold value to segment the image into Blob and background pixel set. Morphological processing is used to adjust the Blob shape after segmentation. Finally, the connectivity analysis and feature extraction of the image are carried out. The number and size of defect features are obtained by fitting the Blob region with a minimum external rectangle, and the recognition of common defects such as glass fiber cloth cleavage, floral jump, hole breaking and stain is realized. The experimental results show that the method is simple and easy to calculate. The detection results are robust and real-time, and it is an effective on-line detection method for fabric defects. Based on the database of SQL and SQL Sever, a glass fiber fabric defect detection software system based on VS2010 platform is developed. The system includes image acquisition module. The man-machine interaction module, image processing module and defect data statistics module have realized the detection of glass fiber cloth defects and fabric grading. Debugging and experimental verification have been carried out on the experimental platform. The results show that the research method in this paper is stable and reliable. Good real-time, meet the expected research and development requirements.
【学位授予单位】:郑州大学
【学位级别】:硕士
【学位授予年份】:2017
【分类号】:TS101.97;TP391.41
【参考文献】
相关期刊论文 前10条
1 王庆海;赵凤霞;李纪峰;金少搏;;基于Blob分析的玻璃纤维织物缺陷检测方法研究[J];郑州大学学报(工学版);2015年06期
2 刘洲峰;赵全军;李春雷;董燕;闫磊;;基于局部统计与整体显著性的织物疵点检测算法[J];纺织学报;2014年11期
3 邵鑫玉;华继钊;;基于机器视觉的无纺布缺陷自动检测系统[J];计算机科学;2014年S1期
4 陈志特;;机器视觉照明光源技术要点分析[J];河南科技;2014年02期
5 刘博超;赵建;孙强;;基于边缘改进的Harris角点检测方法[J];液晶与显示;2013年06期
6 崔玲玲;卢朝阳;李静;李益红;;基于非下采样Contourlet域高斯混合模型的布匹瑕疵识别算法[J];吉林大学学报(工学版);2013年03期
7 朱俊岭;汪军;张孝南;李立轻;陈霞;庞明军;;基于AR模型的机织物线状疵点研究[J];纺织学报;2012年08期
8 祝双武;郝重阳;;基于纹理周期性分析的织物疵点检测方法[J];计算机工程与应用;2012年21期
9 刘源l,
本文编号:1673084
本文链接:https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/1673084.html