硅电池片自动串焊表面缺陷在线视觉检测研究  

Online visual inspection for welding defects of silicon battery chips

作　　者：陈智强 楼佩煌[1] 钱晓明[1] CHEN Zhi-qiang;LOU Pei-huang;QIAN Xiao-ming(College of Mechanical and Electrical Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China)

机构地区：[1]南京航空航天大学机电学院,江苏南京210016

出　　处：《机电工程》2018年第7期773-778,共6页Journal of Mechanical & Electrical Engineering

基　　金：江苏省重点研发计划资助项目(BE2016004-3);江苏省产学研合作前瞻性联合研究资助项目(BY2015003-11);江苏省普通高校研究生科研创新计划资助项目(SJZZ16_0058);南京航空航天大学研究生创新基地(实验室)开放基金项目(KFJJ20170515)

摘　　要：针对太阳能硅电池片集成串焊在线表面缺陷检测存在的问题,提出了基于机器视觉的硅电池片串焊质量在线检测方法,包括特征识别与视觉测量两部分。在特征识别阶段,提出了一种改进的区域生长算法,对断栅特征实现高速稳定地检测,对于焊后赃污划痕缺陷,先将感兴趣区域自适应阈值化,再通过不变矩等特征筛选异类缺陷;在视觉测量方面,利用目标图像对称性,通过若干采样矩形对焊带偏移及片间距进行基于像素点的测量。该方法实现了在每个传送节拍获取片间图像,通过边缘定位并将原始图像分割为多个感兴趣区域并完成高速在线检测的目标。研究结果表明:提出的硅电池片焊后视觉检测方法能够在自动串焊生产线的快节拍下精确测量与识别缺陷,并具有较好的鲁棒性。Aiming at the online quality inspection of welded silicon battery chips,an online quality inspection method of welded silicon battery chips based on visual identification technology was proposed in this paper,including feature recognition and visual measurement. In the feature recognition stage,an improved region growing algorithm was proposed to finish the high-speed and stable detection of gridline-fracture features; For the scratch and dirty features caused by welding,the regions of interest was adaptive thresholded,then the defects were filtered by invariant moments feature. In the field of visual measurement,using the symmetry of target image,pixel based measurements of solder strip offsets and chip spacings were carried out through a number of sampling rectangles. The acquisition of inter chip images at each transmission beat was enabled,the original image was segmented into a number of regions of interest by the edge location method,so that highspeed online detection was completed. The results of actual online test indicate that,the visual inspection method of welded silicon battery chips can identify target defects quickly and accurately,and it has the characteristics of good recognition stability.

关 键 词：自动串焊 视觉测量 在线检测 不变矩特征 区域生长 

分 类 号：TP242.62[自动化与计算机技术—检测技术与自动化装置] TG409[自动化与计算机技术—控制科学与工程]