Journal of Applied Optics, Volume. 41, Issue 2, 327(2020)

Research on detection agorithm of solar cell component defects based on deep neural network

Huaiguang LIU1...2, Anyi LIU1,*, Shiyang ZHOU1,2, Hengyu LIU1 and Jintang YANG1 |Show fewer author(s)
Author Affiliations
  • 1Key Laboratory of Metallurgical Equipment and Control Technology, Wuhan University of Science and Technology, Wuhan 430081, China
  • 2Institute of Robotics and Intelligent Systems, Wuhan University of Science and Technology, Wuhan 430081, China
  • show less
    Figures & Tables(19)
    Principle of photoluminescence imaging
    Camera layout and image view (10×6 component)
    Stitching matches key points of Camera 2
    Single camera projection positioning
    Neighborhood extreme value difference processing of camera Y-direction projection curve
    Edge index of camera
    Movement direction of each camera
    Assembly result of component image
    Single cell plate front and PL image
    Diagrams of corner defects
    Location of corner points to be detected
    Corner screenshot process
    Convolutional neural network models for training
    Part of training samples
    Model training accuracy and loss curve
    • Table 1. Key points required by each camera

      View table
      View in Article

      Table 1. Key points required by each camera

      相机1相机2相机3相机4相机5相机6相机7相机8
      1#边
      2#边
      3#边
      4#边
    • Table 2. Network model structure and some parameters

      View table
      View in Article

      Table 2. Network model structure and some parameters

      model1model2model3
      每层名称类别每层名称类别每层名称类别
      Conv1ConvolutionConv1ConvolutionConv1Convolution
      pooling1Max poolingpooling1Max poolingpooling1Max pooling
      Conv2ConvolutionConv2ConvolutionConv2Convolution
      pooling2Max poolingpooling2Max poolingpooling2Max pooling
      FC5Fully connectionConv3ConvolutionFC5Fully connection
      FC6Fully connectionPooling3Max poolingFC6Fully connection
      FC7Fully connectionFC7Fully connection
      FC8Fully connection
      FC9Fully connection
    • Table 3. Sample images data set

      View table
      View in Article

      Table 3. Sample images data set

      数据集无缺陷第一类 缺陷 第二类 缺陷 黑斑和缺陷 不在角点上 总张数
      训练集8 7008 7008 7008 70034 800
      验证集8008008008003 200
      测试集1 0001 0001 0001 0004 000
    • Table 4. Comparison results of different models recognition accuracy

      View table
      View in Article

      Table 4. Comparison results of different models recognition accuracy

      不同模型第一类 缺陷 第二类 缺陷 黑斑和缺陷 不在角点上 无缺陷识别准确率
      model198.7%95.4%97.8%99.9%97.95%
      model295.7%94.2%78.4%100%92.08%
      model399.5%99.7%97.9%99.9%99.25%
    Tools

    Get Citation

    Copy Citation Text

    Huaiguang LIU, Anyi LIU, Shiyang ZHOU, Hengyu LIU, Jintang YANG. Research on detection agorithm of solar cell component defects based on deep neural network[J]. Journal of Applied Optics, 2020, 41(2): 327

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: OE INFORMATION ACQUISITION AND PROCESSING

    Received: Sep. 23, 2019

    Accepted: --

    Published Online: Apr. 23, 2020

    The Author Email: LIU Anyi (1609399877@qq.com)

    DOI:10.5768/JAO202041.0202006

    Topics