基于LeNet-5模型的太阳能电池板缺陷识别分类
|
摘要
太阳能电池板是光伏发电组件的核心部件,其质量的优劣直接关系安全发电和发电效率。因此,对太阳能电池板进行缺陷检测具有重要的实际价值。考虑到人工检测的低效性和高成本,本文提出利用在深度学习领域图像分类性能良好的卷积神经网络对太阳能电池板图像进行自动识别分类。利用Tensorflow平台Tensorboard的可视化性能,对经典卷积神经网络Le Net-5模型进行结构改善和超参数的调整,并将改进LeNet-5模型与经典LeNet-5模型和支持向量机的分类结果互相对比,结果表明改进LeNet-5模型的分类效果最优。
The solar cell, as the core component of photovoltaic power generation module, its quality directly relates to security and efficiency of power generation. So the defect detection of solar cells plays a very important significant role. Taking into account the inefficiencies and high costs of manual detection, this paper proposes to use the convolutional neural network with good image classification performance in the deep learning field to automatically identify and classify the cell. Based on the visualization performance of Tensorboard on Tensorflow platform, the structure of the classical Le Net-5 model is improved and the hyper parameters are adjusted.Moreover, the classification results of the improved LeNet-5 model are compared with that of the classical LeNet-5 model and support vector machine. The results show that the classification effect of the improved LeNet-5 model is the optimal.
The solar cell, as the core component of photovoltaic power generation module, its quality directly relates to security and efficiency of power generation. So the defect detection of solar cells plays a very important significant role. Taking into account the inefficiencies and high costs of manual detection, this paper proposes to use the convolutional neural network with good image classification performance in the deep learning field to automatically identify and classify the cell. Based on the visualization performance of Tensorboard on Tensorflow platform, the structure of the classical Le Net-5 model is improved and the hyper parameters are adjusted.Moreover, the classification results of the improved LeNet-5 model are compared with that of the classical LeNet-5 model and support vector machine. The results show that the classification effect of the improved LeNet-5 model is the optimal.
引文
[1]人民日报.我国光伏装机容量世界第一[J].创新时代,2017(3):99.DOI:10.3969/j.issn.1674-6538.2017.03.042.People’s Daily.China’s PV installed capacity is the biggest in the world[J].Innovation Time,2017(3):99.DOI:10.3969/j.issn.1674-6538.2017.03.042.
[2]王晗.多晶硅表面缺陷识别及软件检测系统设计与开发[D].上海:上海交通大学,2014:3-11.WANG Han.Polysilicon solar wafer surface defect detection and the design and development of software inspection system[D].Shanghai:Shanghai Jiao Tong University,2014:3-11.
[3]王楠.基于视觉的硅太阳能电池检测方法的研究[D].保定:河北农业大学,2014:11-18.WANG Nan.Research of silicon solar cell based on visual detection method[D].Baoding:Agricultural University of Hebei,2014:11-18.
[4]赵慧阳.基于机器视觉的太阳能电池片表面缺陷检测的研究[D].秦皇岛:燕山大学,2011:11-57.ZHAO Huiyang.Research on surface defect inspection of solar cell based on machine vision[D].Qinhuangdao:Yanshan University,2011:11-57.
[5]李梦园.深度学习算法在表面缺陷识别中的应用研究[D].杭州:浙江工业大学,2015:1-2.LI Mengyuan.Research and application of deep learning algorithm in suface defect recognition[D].Hangzhou:Zhejiang University of Technology,2015:1-2.
[6]冯博.太阳能电池片表面裂纹检测算法的研究与实现[D].长春:吉林大学,2014:32-50.FENG Bo.Research and implementation of crack detection algorithm on the surface of solar cell[D].Changchun:Jilin University,2014:32-50.
[7]LAUER F,SUEN C Y,BLOCH G.A trainable feature extractor for handwritten digit recognition[J].Pattern Recognition,2007,40(6):1816-1824.
[8]LAWRENCE S,GILES C L,TSOI A C,et al.Face recognition:a convolutional neural-network approach[J].IEEE Transactions on Neural Networks,1997,8(1):98-113.
[9]赵志宏,杨绍普,马增强.基于卷积神经网络Le Net-5的车牌字符识别研究[J].系统仿真学报,2010,22(3):638-641.ZHAO Zhihong,YANG Shaopu,MA Zengqiang.License plate character recognition based on convolutional neural network LeNet-5[J].Journal of System Simulation,2010,22(3):638-641.
[10]TIVIVE F H C,BOUZERDOWN A.An eye feature detector based on convolutional neural network[C]//Eighth International Symposium on Signal Processing and ITS Applications.[S.l.]:IEEE,2006:90-93.
[11]SZARVAS M,YOSHIZAWA A,YAMAMOTO M,et al.Pedestrian detection with convolutional neural networks[C]//Intelligent Vehicles Symposium,2005Proceedings.[S.l.]:IEEE,2005:224-229.
[12]SERMANET P,CHINTALA S,LECUN Y.Convolutional neural networks applied to house numbers digit classification[C]//Pattern Recognition(ICPR),21st International Conference on IEEE.[S.l.]:IEEE,2012:3288-3291.
[13]林剑春,杨爱军,沈熠辉.电致发光缺陷检测仪的成像性能评估[J].光学精密工程,2017,25(6):1418-1424.LIN Jianchun,YANG Aijun,SHEN Yihui.Evaluation of imaging performance for electroluminescence defect detector[J].Optics and Precision Engineering,2017,25(6):1418-1424.
[14]施光辉,崔亚楠,刘小娇,等.电致发光(EL)在光伏电池组件缺陷检测中的应用[J].云南师范大学学报(自然科学版),2016,36(2):17-21.SHI Guanghui,CUI Yanan,LIU Xiaojiao,et al.Electrolumninscent application in defects detection of photovoltaic-module[J].Journal of Yunnan Normal University(Natural Sciences Edition),2016,36(2):17-21.
[15]LECUN Y,BENGIO Y,HINTON G.Deep learning[J].Nature,2015,521(7553):436.
[16]LéCUN Y,BOTTOU L,BENGIO Y,et al.Gradient-based learning applied to document recognition[J].Proceedings of the IEEE,1998,86(11):2278-2324.
[17]杨振杰.基于CNN的交通标志识别方法研究[D].天津:天津工业大学,2017:13-15.YANG Zhenjie.Research on traffic sign recognition method based on CNN[D].Tianjin:Tianjin Polytechnic University,2017:13-15.
[2]王晗.多晶硅表面缺陷识别及软件检测系统设计与开发[D].上海:上海交通大学,2014:3-11.WANG Han.Polysilicon solar wafer surface defect detection and the design and development of software inspection system[D].Shanghai:Shanghai Jiao Tong University,2014:3-11.
[3]王楠.基于视觉的硅太阳能电池检测方法的研究[D].保定:河北农业大学,2014:11-18.WANG Nan.Research of silicon solar cell based on visual detection method[D].Baoding:Agricultural University of Hebei,2014:11-18.
[4]赵慧阳.基于机器视觉的太阳能电池片表面缺陷检测的研究[D].秦皇岛:燕山大学,2011:11-57.ZHAO Huiyang.Research on surface defect inspection of solar cell based on machine vision[D].Qinhuangdao:Yanshan University,2011:11-57.
[5]李梦园.深度学习算法在表面缺陷识别中的应用研究[D].杭州:浙江工业大学,2015:1-2.LI Mengyuan.Research and application of deep learning algorithm in suface defect recognition[D].Hangzhou:Zhejiang University of Technology,2015:1-2.
[6]冯博.太阳能电池片表面裂纹检测算法的研究与实现[D].长春:吉林大学,2014:32-50.FENG Bo.Research and implementation of crack detection algorithm on the surface of solar cell[D].Changchun:Jilin University,2014:32-50.
[7]LAUER F,SUEN C Y,BLOCH G.A trainable feature extractor for handwritten digit recognition[J].Pattern Recognition,2007,40(6):1816-1824.
[8]LAWRENCE S,GILES C L,TSOI A C,et al.Face recognition:a convolutional neural-network approach[J].IEEE Transactions on Neural Networks,1997,8(1):98-113.
[9]赵志宏,杨绍普,马增强.基于卷积神经网络Le Net-5的车牌字符识别研究[J].系统仿真学报,2010,22(3):638-641.ZHAO Zhihong,YANG Shaopu,MA Zengqiang.License plate character recognition based on convolutional neural network LeNet-5[J].Journal of System Simulation,2010,22(3):638-641.
[10]TIVIVE F H C,BOUZERDOWN A.An eye feature detector based on convolutional neural network[C]//Eighth International Symposium on Signal Processing and ITS Applications.[S.l.]:IEEE,2006:90-93.
[11]SZARVAS M,YOSHIZAWA A,YAMAMOTO M,et al.Pedestrian detection with convolutional neural networks[C]//Intelligent Vehicles Symposium,2005Proceedings.[S.l.]:IEEE,2005:224-229.
[12]SERMANET P,CHINTALA S,LECUN Y.Convolutional neural networks applied to house numbers digit classification[C]//Pattern Recognition(ICPR),21st International Conference on IEEE.[S.l.]:IEEE,2012:3288-3291.
[13]林剑春,杨爱军,沈熠辉.电致发光缺陷检测仪的成像性能评估[J].光学精密工程,2017,25(6):1418-1424.LIN Jianchun,YANG Aijun,SHEN Yihui.Evaluation of imaging performance for electroluminescence defect detector[J].Optics and Precision Engineering,2017,25(6):1418-1424.
[14]施光辉,崔亚楠,刘小娇,等.电致发光(EL)在光伏电池组件缺陷检测中的应用[J].云南师范大学学报(自然科学版),2016,36(2):17-21.SHI Guanghui,CUI Yanan,LIU Xiaojiao,et al.Electrolumninscent application in defects detection of photovoltaic-module[J].Journal of Yunnan Normal University(Natural Sciences Edition),2016,36(2):17-21.
[15]LECUN Y,BENGIO Y,HINTON G.Deep learning[J].Nature,2015,521(7553):436.
[16]LéCUN Y,BOTTOU L,BENGIO Y,et al.Gradient-based learning applied to document recognition[J].Proceedings of the IEEE,1998,86(11):2278-2324.
[17]杨振杰.基于CNN的交通标志识别方法研究[D].天津:天津工业大学,2017:13-15.YANG Zhenjie.Research on traffic sign recognition method based on CNN[D].Tianjin:Tianjin Polytechnic University,2017:13-15.