摘要
为了解决在丝网印刷流水线末端光伏太阳能晶硅电池的自动缺陷检测和颜分类的问题，通过对太阳能电池的检测方法进行分析，提出了基于机器视觉的电池片缺陷检测及颜分选的解决方案，开发了光伏太阳能晶硅电池片检测系统。
首先介绍了太阳能电池的制备工艺，分析了太阳能电池的常见缺陷种类和系等级及其形成原因，并提出了相应的检测标准和要求。同时进行光伏太阳能晶硅电池片检测系统整体方案设计，分别对系统硬件和系统软件进行分析设计，然后根据检测系统要求完成了工业相机，镜头，灰度卡等核心元件的选型以及检测功能模块化，离线在线相结合的软件系统设计。
研究了太阳能晶硅电池片缺陷检测算法，介绍了电池片图像的彩校正，区域提取，硅片定位，工艺点屏蔽等预处理过程。提出了基于亚像素的电池片的尺寸测量方法。针对破损缺陷，分别使用形态学和参考模板的检测方法进行实验，通过比较分析选择合适的检测算法。针对栅线印刷缺陷，根据栅线分布特征将检测步骤细分为栅线提取，细栅检测，主栅检测。针对脏污缺陷，提取出基于改进的局部阈值分割方法。
研究了太阳能晶硅电池片颜分选算法，介绍了常用的颜空间及其转化方法，并使用HSI通道进行颜直方图特征提取。分析了传统的颜分选算法后，提出了一种基于神经网络的颜分选算法，并通过实验对比分析两种方法的运行效率和精确度，验证了本算法的优越性。
最后，针对本课题的太阳能晶硅电池片缺陷检测及颜分选系统，在线检测与人工目检相结合，分别从系统的精确度，高效性以及稳定性进行综合性能的实验分析。实验数据表明，系统的综合性能可以满足实际生产需求。
关键词：机器视觉；检测系统；缺陷检测；颜分选；神经网络
Abstract
In order to solve the problem of automatic defect detection and color classification of silicon cells at the end of screen printing pipeline,through the analysis of solar cell detection methods,a solution based on machine vision for defect detection and color sorting of cells is proposed,and developing an inspection system of photovoltaic solar crystalline silicon cell.sorting out
Firstly,analyzing the common defect types,color grades and their causes of formation of solar cells by introducing the preparation process of solar cells,afterwards putting forward the corresponding detection standards and requirements.At the same time the overall scheme design of photovoltaic solar crystal silicon cell detection system is carried out,and the system hardware and system software are analyzed and designed respectively.Then,the selection of core components,such as industrial camera,lens and gray card,and offline online software system design with detection function modular a
re completed according to the requirements of the detection system.
secondly,the defect detection algorithm of solar crystalline silicon cells is discussed.First, introducing the preprocessing processes including color correction,region extraction,silicon wafer positioning,process point shielding and so on.After that,proposing a subpixel-based cell size measurement method.Aiming at the damage defects,the detection methods of morphology and reference template are used to carry out experiments,and the appropriate detection algorithm is selected through comparative analysis.Aiming at the defects of grid line printing,the detection steps are subdivided into grid line extraction,fingers detection and busbars detection according to the distribution characteristics of grid line.Aiming at the dirt defect,an improved local threshold segmentation method was extracted.
Then,research on the color sorting algorithm of solar crystalline silicon cells.First the color histogram feature extraction is carried out using HSI channel through introducing the common color space and its transformation method.Then a color sorting algorithm based on neural network is advanced with the traditional color sorting algorithm analyzed,and the operation efficiency and accuracy of the two methods are compared and analyzed experimentally,which proves the superiority of the algorithm.
Finally,aiming at the defect detection and color sorting system of solar crystal silicon cells in this topic,the comprehensive performance of the system was analyzed from the perspective of accuracy，efficiency and stability combined with online detection and manual eye inspection.Experimental data show that the comprehensive performance of the system can meet the actual production requirements.
Key Words:Machine Vision;Detection System;Defect Detection;Color Sorting;Neural Network
目录
摘要..................................................................................................................................... I Abstract ............................................................................................................................. I I 第一章绪论. (1)
1.1 研究背景 (1)
1.2 国内外研究现状 (2)
1.2.1 机器视觉研究现状 (2)
1.2.2 太阳能电池检测的研究现状 (3)
1.3 本文研究目的与意义 (5)
1.4 本文研究内容和组织结构 (6)
1.4.1 研究内容 (6)
1.4.2 组织结构 (6)
第二章缺陷检测与颜分选系统的分析与设计 (8)
2.1 引言 (8)
2.2 电池片视觉检测关键问题 (8)
2.2.1 太阳能电池制备工艺 (8)
2.2.2 电池片缺陷与系 (9)
2.2.3 检测标准与要求 (12)
2.3 检测系统整体方案设计 (13)
2.4 检测系统硬件选型与设计 (14)
2.4.1 工业相机选型 (15)
2.4.2 相机镜头选型 (17)
2.4.3 灰度卡的选型 (19)
2.4.4 其他硬件 (20)
2.5 检测系统软件设计 (21)
2.6 本章小结 (23)
第三章电池片缺陷检测算法研究 (24)
3.1 引言 (24)
3.2 电池片图像预处理 (24)
3.2.1 图像的白平衡彩校正算法 (24)
3.2.2太阳能晶硅电池片区域提取 (26)
3.2.3基于模板匹配的电池片定位 (27)
3.2.4基于仿射变换的工艺点屏蔽 (30)
3.3基于亚像素的电池片尺寸测量 (32)
3.4电池片破损缺陷检测算法研究 (37)
3.4.1基于形态学检测方法 (37)
3.4.2基于模板检测方法 (39)
3.4.3破损检测实验分析 (40)
3.5电池片栅线缺陷检测算法研究 (40)
3.5.1电池片栅线提取 (40)
3.5.2电池片细栅检测 (41)
3.5.3电池片主栅检测 (43)
3.6电池片脏污缺陷检测算法研究 (44)
3.7本章小结 (46)
第四章电池片颜分选算法研究 (47)
4.1引言 (47)
4.2颜空间变换与特征提取 (47)
4.3电池片颜分选算法 (50)
4.3.1基于相似度和距离的颜分选算法 (50)
4.3.2基于神经网络的颜分选算法 (51)
4.3.3颜分选算法的对比分析 (57)
4.4本章小结 (58)
第五章检测系统性能分析 (59)
5.1引言 (59)
5.2检测系统精确度实验 (60)
5.3检测系统高效性实验 (60)
5.4检测系统稳定性实验 (61)
5.5本章小结 (62)
总结与展望 (63)
研究成果 (63)

版权声明：本站内容均来自互联网，仅供演示用，请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系QQ:729038198，我们将在24小时内删除。