| 本文已被：浏览 828次 下载 280次 |
码上扫一扫！ |
|
|
| 基于视觉的动力电池焊后质量检测 |
|
吴博尧, 秦磊, 张庆茂, 马琼雄1,2,3
|
1. 华南师范大学 广州市特种光纤光子器件与应用重点实验室, 广州 510006;2.
2. 华南师范大学 广东省微纳光子功能材料与器件重点实验室, 广州 510006;3.
3. 深圳市联赢激光股份有限公司, 深圳 518055
|
|
| 摘要: |
| 为了对动力电池的焊后质量进行检测，针对检测时对比度低、背景复杂和干扰等问题，提出一种融合了动态阈值与全局阈值、行程处理和主成分分析-支持向量机（PAC-SVM）分类模型的焊后质量检测方法.首先，提出了一种结合动态阈值和全局阈值的混合阈值算法来分割焊缝和瑕疵；其次，利用形态学和行程处理来消除焊缝周边干扰得到更真实的焊缝边缘；最后，融合灰度特征、几何特征和矩特征建立7维特征向量的针孔模型，并采用带主成分分析的支持向量机检测针孔.结果表明，采用文中提出的方法，可获得良好的检测效果. |
| 关键词: 混合阈值 行程处理 主成分分析 支持向量机 |
| DOI：10.12073/j.hjxb.2018390237 |
| 分类号:TG409 |
| 基金项目:华南师范大学青年教师科研培育基金（15KJ13）；广东省科技项目（2013B090600045、2013B090200003、2014B010131004、2014B010124002、2014B090903014、2015B090920003、2016B090917002、2016B090926004）；广东省自然科学基金（2016A030313456） |
|
| Research on vision-based post-welding quality inspection of power battery |
|
WU Boyao, QIN Lei, ZHANG Qingmao, MA Qiongxiong1,2,3
|
1. Guangzhou Key Laboratory for Special Fiber Photonic Devices and Applications, South China Normal University, Guangzhou 510006, China;2.
2. Guangdong Provincial Key Laboratory of Nanophotonic Functional Materials and Devices, South China Normal University, Guangzhou 510006, China;3.
3. United Winners Laser Co., Ltd., Shenzhen 518055, China
|
| Abstract: |
| In order to inspect the post-welding quality of the power battery, this paper proposes a post-weld quality inspection method. This method combines the dynamic threshold and the global threshold, the runs processing and the PCA-SVM classification model for the problems of low contrast, complex background and interference. Firstly, a hybrid threshold algorithm combining dynamic threshold and global threshold is proposed to segment weld seam and defects. Secondly, get a more real edge of the weld seam though use morphology and runs processing to eliminate the interference around the weld seam; Finally, a 7-dimensional feature vector is designed from three aspects:gray features, geometric features and moments. The support vector machine model with principal component analysis is used to inspect pinholes. The results show that the proposed method can achieve good inspection quality. |
| Key words: hybrid threshold runs processing principal component analysis support vector machine |
