Integrated Enhanced Feature Fusion Module For Photovoltaic Defect Detection Network

To address the challenges of high similarity between surface defect features and background features, as well as the variability in geometric shapes in solar cell surface defects, a solar cell surface defect detection network model called EMFAnet is proposed, which combines multi-scale features and attention mechanisms. Firstly, a Three-path Coordinate Attention (TCA) is designed to enhance the detection capability in the interested regions; secondly, to improve the accuracy and precision of defect localization and description, TCA is fused with deformable convolutions to form a Multi-Path Coordinate-Deformable Conv Module (MCD); lastly, to adapt to complex backgrounds and effectively detect defects at various scales, an enhanced multi-scale feature fusion method is designed. Experimental results demonstrate that the proposed approach achieves significantly higher accuracy in identifying solar cell EL defects compared to existing state-of-the-art detection algorithms.