基于机器视觉的太阳能电池外观检测技术研究

发布时间：2018-01-11 13:02

本文关键词：基于机器视觉的太阳能电池外观检测技术研究　出处：《苏州大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着能源需求的不断增长,太阳能作为一种清洁环保的可再生资源,逐渐开始被广泛利用。太阳能电池在制作过程中通常需要通过严格的质量检测,如是否存在物理破损、表面划痕,是否存在断栅,印刷图案是否清晰正确等问题。传统工业模式下,这些检测工序需要检测人员人工目测,人工检测不仅检测效率低下而且容易出现因视觉疲劳和主观臆断造成检测错误情况。近年来随着计算机技术和数字图像处理技术的快速发展,使得利用机器视觉对太阳能电池进行外观缺陷检测成为现实。本文首先分析了一般检测系统中使用简单阈值法进行图像分割的缺点,由于亮度不均以及缺陷区域对比度不高,使用单一阈值往往容易造成图像分割错误。针对此问题,通过结合局部自适应阈值处理法和区域生长法的优点,提出了一种更为有效的图像分割方法。实验结果验证新方法对亮度不均具有较好的鲁棒性,并且能够准确的提取隐藏型的外观缺陷。太阳能电池组件通常由多个太阳能电池片通过串联焊接在一起,因此相机标定的精度对于太阳能视觉检测系统也尤为重要。但是用于图像采集的广角镜头存在桶形畸变。本文采用基于线性插值的畸变校准方法,并利用最小二乘法对内部栅线进行直线拟合,较少了镜头畸变造成的坐标误差。最后,本文在Windows平台上实现了太阳能外观视觉检测系统,程序测试结果表明本文所述缺陷检测方法和坐标标定方法均具有较好的可行性和较高的准确度。
[Abstract]:With the growing demand for energy, solar energy as a renewable resource, gradually began to be widely used in the production process. The solar battery is usually required by strict quality testing, such as the presence of physical damage, surface scratches, the presence of broken gate, the printing pattern is clear and correct. The traditional industrial mode. These detection procedures need to detect artificial visual inspection, manual testing is not only low detection efficiency and prone to visual fatigue and subjective cause detection error. In recent years, with the rapid development of computer technology and digital image processing technology, the solar cell appearance defect detection based on machine vision become a reality. This paper firstly analyzes the general detection system use a simple threshold method of defect image segmentation, because of uneven brightness and defect area of High contrast, using a single threshold image segmentation is often likely to cause errors. To solve this problem, by combining the advantages of local adaptive thresholding method and region growing method, put forward a more effective method in image segmentation. The experimental results verify that the new method of uneven brightness has better robustness, and can look accurate extraction of hidden defects. Solar cell components are usually composed of a plurality of solar cells by welded together in series, so the precision of camera calibration for the solar visual detection system is also very important. But the existence of barrel distortion of wide-angle lens for image acquisition. This paper uses distortion calibration method based on linear interpolation, and linear fitting the internal gate line by using least squares method, less error caused by coordinate lens distortion. Finally, this paper implements solar appearance on the Windows platform The result of the program test shows that the defect detection method and the coordinate calibration method described in this paper have good feasibility and high accuracy.

【学位授予单位】：苏州大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41;TM914.4

【相似文献】