CASIA OpenIR  > 毕业生  > 硕士学位论文
基于显微视觉的手机白玻表面缺陷检测方法研究
袁伦喜
Subtype工程硕士
Thesis Advisor张正涛
2017-05
Degree Grantor中国科学院研究生院
Place of Conferral北京
Keyword缺陷分割 缺陷分类 视觉显著性 视觉测量 手机白玻
Abstract摘要
手机白玻是智能手机的一个重要部件。随着智能手机的日渐普及,白玻的用量及产量也迅猛增长。白玻表面质量直接影响手机的外观和销售。传统的白玻表面缺陷检测主要依赖于有经验的工人,但由于白玻市场需求的增大、企业竞争的加剧以及客户对白玻质量要求的提高,人工检测方法再难满足企业对白玻品质检测的要求。本文针对白玻表面缺陷检测这一问题,研制了一套基于显微视觉的手机白玻表面缺陷检测装置,对手机白玻表面划伤的分割、分类和融合方法以及白玻边缘崩边的检测方法等内容进行研究。本文的主要研究工作包括以下几个方面:
第一,针对白玻表面缺陷检测这一问题,本文搭建了一套基于显微视觉的手机白玻表面缺陷检测装置,并研制了一套手机白玻表面缺陷检测软件。该装置主要包括显微视觉装置和精密传送装置两个子装置。通过合理设置相关硬件参数,该装置能够有效提取多种类型缺陷的高质量图像。
第二,针对白玻图像中划伤分割问题,本文提出了一种基于迭代阈值的显著性图分割算法。在这一算法中,考虑到划伤缺陷对比度较低,首先设计了基于直方图对比度的显著性算法来增强图像对比度;然后,采用迭代阈值法来分割显著性增强后的白玻图像。实验证明,该算法能够实现低对比度划伤缺陷的有效分割,并具有较好的实时性能和抗干扰能力。
第三,由于划伤没有固定形状,并且有些划伤是断续的或散乱分布的,这导致直接测量划伤的尺寸存在困难。针对这一问题,首先提取了划伤的特征,并对划伤采用支持向量机进行子类区分;在此基础上,提出了一种基于区域生长的划伤融合方法,对不同子类的划伤采用不同的策略进行生长融合;最后,提取生长融合后的划伤轮廓的最小外接矩形来表征划伤,并用最小外接矩形的尺寸来近似划伤尺寸。实验表明,该方法能够适用于多种不规则形状划伤的尺寸测量,提高了划伤尺寸的测量准确度。
第四,为检测白玻边缘的崩边缺陷,提出了一种崩边缺陷的提取与测量算法。在该算法中,首先,利用0-1阶跃法精确定位白玻边缘,测量白玻边缘宽度,在测量过程中,采用随机采样一致性来拟合白玻竖直边缘,测量白玻倾斜角度,以校正因图像倾斜导致的测量误差。然后,利用波峰截取法提取白玻水平和竖直边缘处的崩边。最后,根据圆弧边缘的曲率改进波峰截取法,利用改进后算法提取白玻圆弧处崩边,以适应因圆弧导致的边缘宽度变化。实验表明,本文提出的崩边提取与测量算法能够实现对较小崩边缺陷的准确定位,并且具有较高的实时性。
Other AbstractWhite glass cover is one of the most significant components of mobile phone. Its quality directly affects the appearance of mobile phone, thus affecting phone’s sales. Traditional defect detection of the white glass cover is mainly dependent on experienced workers. Faced to the change of the market, such as more fierce competition, increased demand and a higher requirement of quality, the manual detection method cannot satisfy the needs of industrial applications. This thesis focuses on the problem of automatic defect detection of white glass cover and a detection system is developed based on micro vision. Several related issues are proposed and investigated, including the segmentation, classification and fusion of the scratch and the crack detection for the edge.The main contributions are as follows:
Firstly, focusing on the problem of automatic defect detection of white glass cover, a detection system is developed based on micro vision. Two subsystems are included, which respectively are microscopic visual device and precision transmission device. By properly setting the parameters of the light source and camera, high quality images for many types of the defects could be effectively achieved.
Secondly, a method is proposed to segment the scratches in white glass images, in which the low contrast between the scratch and surrounding background is taken into consideration. In the method, the HC algorithm is employed to enhance the contrast, and an iterative threshold algorithm is proposed to segment scratches in the enhanced image. The method could effectively segment the scratch with improved real-time performance. Moreover, this method has certain anti-interference ability.
Thirdly, due to unfixed shape, it is hard to measure the size of scratch directly. In addition, some scratches are intermittent and scattered, which makes the measurement of the scratch even harder. To solve this problem, we firstly extracted the features of scratches and then the SVM is utilized to determine the subclass of scratches. Subsequently, a scratch fusion algorithm is proposed based on region growing to connect intermittent and scattered scratches. Finally, the minimum circumscribed rectangle fitting scratch is obtained to approximate the size of the scratch. Improved accuracy has been achieved when applying the method to the measurement of scratches with irregular shapes.
Finally, for the detection of the collapse defect, a collapses extraction and measurement method is proposed. In the method, we firstly use the 0-1 step algorithm to locate the edge of white glass precisely, and then the width of the edge is obtained. During the process, the RANSAC algorithm is employed to fit the vertical edge, and the titled angle is determined to modify the error of the measured width, which is caused by the skew of the image. Consequently, the Wave Peak Crop algorithm is proposed to detect the collapses on vertical edge. Furthermore, the algorithm is improved by the curvature when detecting the collapses on arc edge. The collapses are determined with the improved algorithm. Experiments demonstrate that, the collapses, even those with small sizes, could be effectively located with satisfactory accuracy and a high real-time capacity.
Document Type学位论文
Identifierhttp://ir.ia.ac.cn/handle/173211/14749
Collection毕业生_硕士学位论文
Recommended Citation
GB/T 7714
袁伦喜. 基于显微视觉的手机白玻表面缺陷检测方法研究[D]. 北京. 中国科学院研究生院,2017.
Files in This Item:
File Name/Size DocType Version Access License
基于显微视觉的手机白玻表面缺陷检测方法研(3432KB)学位论文 暂不开放CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[袁伦喜]'s Articles
Baidu academic
Similar articles in Baidu academic
[袁伦喜]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[袁伦喜]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.