This paper presents a benchmark dataset and results for automatic detection and classification using deep learning models trained on 24 defects and features in EL images of crystalline silicon solar cells. The dataset consists of 593 cell images with ground truth masks corresponding to the pixel-level labels for each feature and defect.
Visualizing feature map (The figure illustrates the change in the feature map after the SRE module.) We propose a photovoltaic cell defect detection model capable of extracting topological knowledge, aggregating local multi-order dynamic contexts, and effectively capturing diverse defect features, particularly for small flaws.
The models tested are effective in detecting, localizing, and quantifying multiple features and defects in EL images of solar cells. These models can thus be used to not only detect the presence of defects, but to track their evolution over time as modules are re-imaged throughout their lifetime.
Automatic defect detection and classification in solar cells is the subject of many publications since EL imaging of silicon solar cells was first introduced by Fuyuki et al. for detection of deteriorated areas in solar cells in 2005.
The convolution-based attention mechanism in MSCA effectively aggregates the texture structures of local defects and differentiates between pixel points, making it particularly adept at detecting less conspicuous photovoltaic cell defects.
Abstract: A solar cell defect detection method with an improved YOLO v5 algorithm is proposed for the characteristics of the complex solar cell image background, variable defect morphology, and large-scale differences.