Intelligent defect detection in electroluminescence images of photovoltaic modules using MobileNetV2

With the rapid growth of the photovoltaic (PV) industry, fast and accurate defect-detection techniques are becoming increasingly important. Manual inspection of PV modules using electroluminescence (EL) imaging is time-consuming and prone to errors. This study proposes a clever method for detecting defects using a lightweight deep learning model based on the MobileNetV2 architecture. The model learns from a dataset of EL images showing two common types of defects: cracks and dark areas. It also contains defect-free cells. To improve robustness to typical EL acquisition variability, an EL-tailored data augmentation pipeline is applied, including geometric transformations and photometric adjustments (brightness and contrast). During testing, it takes only 0.913 seconds to predict an image. This demonstrates a good compromise between speed and accuracy. This approach offers a promising solution for low-cost, near-real-time quality inspection of photovoltaic modules using artificial intelligence.