Open AccessMicro-cracks Detection of Polycrystalline Solar Cells with Transfer Learning
Author(s) -
Nawei Zhang,
Shuo Shan,
Haikun Wei,
Kanjian Zhang
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1651/1/012118
Subject(s) - monocrystalline silicon , crystallite , photovoltaic system , domain (mathematical analysis) , transfer of learning , materials science , computer science , artificial intelligence , principal (computer security) , biological system , optoelectronics , electronic engineering , engineering , electrical engineering , mathematics , mathematical analysis , silicon , metallurgy , biology , operating system
As the photovoltaic (PV) systems are universally utilized in power systems, the defect of solar cells, the core components of PV system requires to be detected in a low-cost and high-efficiency method, because of its wide application and direct influence on power generation efficiency and system stability. The principal aim of this study is to apply transfer learning in defect detection of polycrystalline cells with image-level labelled mono images and unlabelled poly images. The experiment was conducted with DAN models based on a MK-MMD to represent the distance between source domain and target domain and Resnet-50 backbone to learn transferable features. Three contrast methods were designed to make comparison with DAN. It has been proved that DAN model based on MK-MMD and Resnet-50 shows a promising performance on transferring from monocrystalline tasks to polycrystalline tasks.