基于图像的PCB板缺陷检测技术及应用

发布时间：2018-03-14 09:28

本文选题：印刷电路板　切入点：缺陷检测　出处：《重庆理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：当今,电子工业在国家经济发展中扮演着越来越重要的角色,作为各种电子元器件的高度信息集合,印刷电路板(PCB)被广泛应用在电子工业中的各个领域。经济的不断发展促使电子技术不断地提高,轻薄、便捷的电子工艺成为潮流,高密集、高集成度成为PCB的发展趋势,这给传统的PCB质检带来了十分巨大的挑战。传统的人工检测方法存在着速度慢、时间长和易漏检等等问题,完全无法适应技术和工艺的快速发展,怎样实现精准、高效的PCB自动缺陷检测,一直是电子工业领域非常重视的一个问题。同时,个人和中小企业对检测PCB缺陷的需求越来越高,能否达到低成本、高精度是首要考虑的问题。因此,研究如何通过低成本的图像处理技术提高PCB板缺陷检测的精度具有重要的研究意义。基于图像的PCB缺陷检测流程包括图像预处理、图像配准、图像分割和图像识别等方面,其中图像预处理又包含图像增强、图像平滑和图像锐化操作。根据以上流程和检测的需要,论文的三个主要工作及研究内容如下:1.对比分析得到适当算法。为了在图像配准前和图像分割后得到较为理想的目标处理图像,本文在现有算法的基础上进行算法的对比分析,最终选取灰度变换,自适应滤波,梯度算子作为预处理流程的处理算法;选取最大类间方差阀值作为图像分割操作的算法。分析表明,得到的适当算法有利于PCB图像的配准和识别的工作。2.改进RHT提高配准精度。在图像配准操作上,针对传统算法计算量大,耗时较多的问题,本文提出了一种结合主成分分析(PCA)和分段随机霍夫变换的PCB板圆Mark定位的算法。该算法首先对载入原始彩色PCB板图像进行灰度二值化后采用Canny算子提取图像边缘,然后去掉所得到图像中的交叉点和小线段,并标记剩下的线段,找出大于所设阀值的线段,再运用PCA分析降维去噪,并保留类圆的曲线段,之后采用分段RHT分析得到圆的个数及相关参数,最后结合以上得到的圆参数,通过最小二乘拟合得到所要的圆Mark。实验表明:相比与传统的模版匹配和随机霍夫变换,该算法有效地提高了圆Mark的识别精度和定位精度。3.设计系统识别缺陷。利用MATLAB的图像工具库和上述算法的实现和改进原理,通过MATLAB的GUI功能编制PCB缺陷检测软件。首先分析了PCB缺陷检测软件的处理流程和缺陷识别方法,然后对编制的软件系统进行了功能展示,最后通过缺陷检测实例测试了软件分别在检测短路断路和毛刺缺损的识别效果。实验表明:本文实现的软件系统能精确地检测出PCB板的短路、断路、缺损、毛刺等缺陷。
[Abstract]:Nowadays, the electronic industry plays a more and more important role in the national economic development. Printed circuit board (PCB) has been widely used in various fields of electronic industry. With the development of economy, electronic technology has been continuously improved, thin and convenient electronic technology has become the trend, high density and high integration become the development trend of PCB. This brings a great challenge to the traditional PCB quality inspection. The traditional manual inspection methods have some problems, such as slow speed, long time and easy to miss the inspection, so they can not adapt to the rapid development of technology and technology, and how to achieve precision. Efficient PCB automatic defect detection has always been a very important problem in the electronic industry. At the same time, the demand of individuals and small and medium-sized enterprises to detect PCB defects is increasing. Whether or not to achieve low cost and high precision is the first consideration. It is of great significance to study how to improve the accuracy of PCB board defect detection by low cost image processing technology. The PCB defect detection flow based on image includes image preprocessing, image registration, image segmentation and image recognition. Image preprocessing includes image enhancement, image smoothing and image sharpening. The three main work and research contents of this paper are as follows: 1.Contrastive analysis of the appropriate algorithm. In order to get a better target processing image before image registration and after image segmentation, this paper carries out a comparative analysis of the algorithms based on the existing algorithms. Finally, grayscale transform, adaptive filter and gradient operator are selected as preprocessing algorithms, and the maximum inter-class variance threshold is chosen as the algorithm for image segmentation. The appropriate algorithm is propitious to the work of PCB image registration and recognition. The improvement of RHT improves the registration accuracy. In the image registration operation, the traditional algorithm has a lot of computation and time consuming. In this paper, an algorithm of PCB circular Mark location based on principal component analysis (PCA) and piecewise random Hough transform is proposed. Firstly, the image edge is extracted by using Canny operator after binarization of the gray scale of the loaded original color PCB board image. Then the intersection points and small line segments in the image are removed, and the remaining line segments are marked to find the lines larger than the threshold, and then to use PCA analysis to reduce the dimension and noise, and to preserve the curve segments of the circle. Then, the number of circles and the relative parameters are obtained by piecewise RHT analysis. Finally, the circle Mark is obtained by least square fitting combined with the above circle parameters. The experimental results show that compared with the traditional template matching and random Hough transform, The algorithm effectively improves the recognition accuracy and positioning accuracy of circular Mark. 3. The design system identifies defects. Using the image tool library of MATLAB and the realization and improvement principle of the above algorithm, The GUI function of MATLAB is used to compile the PCB defect detection software. Firstly, the processing flow and defect identification method of PCB defect detection software are analyzed, and then the function of the software system is demonstrated. Finally, the effect of the software in detecting short circuit break and burr defect is tested by an example of defect detection. The experiment shows that the software system can accurately detect the short circuit, open circuit, defect, burr and other defects of PCB board.
【学位授予单位】：重庆理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TN41

【参考文献】