Optimizing feature extraction and fusion for high-resolution defect detection in solar cells

In this paper, we propose a novel architecture for defect detection in electroluminescent images of polycrystalline silicon solar cells, addressing the challenges posed by subtle and dispersed defects. Our model, based on a modified Swin Transformer, incorporates key innovations that enhance feature...

Full description

Saved in:
Bibliographic Details
Main Authors: Hoanh Nguyen, Tuan Anh Nguyen, Nguyen Duc Toan
Format: Article
Language:English
Published: Elsevier 2024-12-01
Series:Intelligent Systems with Applications
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2667305324001170
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850244707380101120
author Hoanh Nguyen
Tuan Anh Nguyen
Nguyen Duc Toan
author_facet Hoanh Nguyen
Tuan Anh Nguyen
Nguyen Duc Toan
author_sort Hoanh Nguyen
collection DOAJ
description In this paper, we propose a novel architecture for defect detection in electroluminescent images of polycrystalline silicon solar cells, addressing the challenges posed by subtle and dispersed defects. Our model, based on a modified Swin Transformer, incorporates key innovations that enhance feature extraction and fusion. We replace the conventional self-attention mechanism with a novel group self-attention mechanism, increasing the mAP50:5:95 score from 50.12 % to 52.98 % while reducing inference time from 74 ms to 62 ms. We also introduce a spatial displacement with shift convolution module, replacing the traditional Multi-Layer Perceptron, which further enhances the model's receptive field and improves precision and recall. Additionally, our fast multi-scale feature fusion mechanism effectively combines high-resolution details with high-level semantic features from different network layers, optimizing defect detection accuracy. Experimental results on the PVEL-AD dataset demonstrate that our model achieves the highest mAP50 score of 83.11 % and an F1-Score of 84.33 %, surpassing state-of-the-art models while maintaining a competitive inference time of 66.3 ms. These findings highlight the effectiveness of our innovations in improving defect detection accuracy and computational efficiency, making our model a robust solution for quality assurance in solar cell manufacturing.
format Article
id doaj-art-e621ed40a9744e328d85bbdbbe509fba
institution OA Journals
issn 2667-3053
language English
publishDate 2024-12-01
publisher Elsevier
record_format Article
series Intelligent Systems with Applications
spelling doaj-art-e621ed40a9744e328d85bbdbbe509fba2025-08-20T01:59:39ZengElsevierIntelligent Systems with Applications2667-30532024-12-012420044310.1016/j.iswa.2024.200443Optimizing feature extraction and fusion for high-resolution defect detection in solar cellsHoanh Nguyen0Tuan Anh Nguyen1Nguyen Duc Toan2Corresponding author.; Faculty of Electrical Engineering Technology, Industrial University of Ho Chi Minh City, Ho Chi Minh City 700000, VietnamFaculty of Electrical Engineering Technology, Industrial University of Ho Chi Minh City, Ho Chi Minh City 700000, VietnamFaculty of Electrical Engineering Technology, Industrial University of Ho Chi Minh City, Ho Chi Minh City 700000, VietnamIn this paper, we propose a novel architecture for defect detection in electroluminescent images of polycrystalline silicon solar cells, addressing the challenges posed by subtle and dispersed defects. Our model, based on a modified Swin Transformer, incorporates key innovations that enhance feature extraction and fusion. We replace the conventional self-attention mechanism with a novel group self-attention mechanism, increasing the mAP50:5:95 score from 50.12 % to 52.98 % while reducing inference time from 74 ms to 62 ms. We also introduce a spatial displacement with shift convolution module, replacing the traditional Multi-Layer Perceptron, which further enhances the model's receptive field and improves precision and recall. Additionally, our fast multi-scale feature fusion mechanism effectively combines high-resolution details with high-level semantic features from different network layers, optimizing defect detection accuracy. Experimental results on the PVEL-AD dataset demonstrate that our model achieves the highest mAP50 score of 83.11 % and an F1-Score of 84.33 %, surpassing state-of-the-art models while maintaining a competitive inference time of 66.3 ms. These findings highlight the effectiveness of our innovations in improving defect detection accuracy and computational efficiency, making our model a robust solution for quality assurance in solar cell manufacturing.http://www.sciencedirect.com/science/article/pii/S2667305324001170Feature extractionFeature fusionSwin transformerDefect detectionElectroluminescent images
spellingShingle Hoanh Nguyen
Tuan Anh Nguyen
Nguyen Duc Toan
Optimizing feature extraction and fusion for high-resolution defect detection in solar cells
Intelligent Systems with Applications
Feature extraction
Feature fusion
Swin transformer
Defect detection
Electroluminescent images
title Optimizing feature extraction and fusion for high-resolution defect detection in solar cells
title_full Optimizing feature extraction and fusion for high-resolution defect detection in solar cells
title_fullStr Optimizing feature extraction and fusion for high-resolution defect detection in solar cells
title_full_unstemmed Optimizing feature extraction and fusion for high-resolution defect detection in solar cells
title_short Optimizing feature extraction and fusion for high-resolution defect detection in solar cells
title_sort optimizing feature extraction and fusion for high resolution defect detection in solar cells
topic Feature extraction
Feature fusion
Swin transformer
Defect detection
Electroluminescent images
url http://www.sciencedirect.com/science/article/pii/S2667305324001170
work_keys_str_mv AT hoanhnguyen optimizingfeatureextractionandfusionforhighresolutiondefectdetectioninsolarcells
AT tuananhnguyen optimizingfeatureextractionandfusionforhighresolutiondefectdetectioninsolarcells
AT nguyenductoan optimizingfeatureextractionandfusionforhighresolutiondefectdetectioninsolarcells