Multi-scale YOLOv5 for solar cell defect detection

被引：0

作者：

Chen Y. ^{[1
]}

Liao F. ^{[1
]}

Huany X. ^{[1
]}

Yang J. ^{[2
]}

Gong H. ^{[1
]}

机构：

[1] College of Science, Chongqing University of Technology, Chongqing

[2] Sichuan YC Garden Technology Co. ，Ltd, Yibin

来源：

Guangxue Jingmi Gongcheng/Optics and Precision Engineering | 2023年 / 31卷 / 12期

关键词：

attention networks; defect detection; deformable convolution v2; solar cells; YOLOv5;

D O I：

10.37188/OPE.20233112.1804

中图分类号：

学科分类号：

摘要：

Herein，to realize high-precision crack and break defect detection in solar cells under electrolu⁃ minescent（EL）conditions，the multi-scale You Only Look Once version 5（YOLOv5）model is used for solar-cell defect detection under real industrial conditions. First，an improved feature-extraction network combining deformable convolution version 2（DCNv2）and coordinate attention（CA）is proposed to wid⁃ en the receptive field of small target defects and enhance the extraction of small-scale defect features. Sec⁃ ond，an improved path aggregation network（PANet），called CA-PANet，is proposed for integrating the CA and cross-layer cascade in a path aggregation network to multiplex shallow features. Notably，the CA-PANet combines deep and shallow features to enhance the feature fusion of defects at different scales，im⁃ prove the feature representation of defects，and increase the defect detection accuracy. The low computa⁃ tional cost of the lightweight CA ensures the real-time performance of the model. Experimental results in⁃ dicate that the mean average precision（mAP）of the YOLOv5 model combining DCNv2 and CA can reach 95. 4%，which is 3% higher than that of the YOLOv5 model and 1. 4% higher than that of the YOLOX model. The improved YOLOv5 model can achieve a frame rate of up to 51 frames per second（FPS），meeting industrial real-time requirements. Compared with other algorithms，the improved YOLOv5 mod⁃ el can accurately detect cracks and break defects in EL solar cells，satisfying the demand for real-time，high-precision defect detection under industrial conditions in photovoltaic power plants. © 2023 Chinese Academy of Sciences. All rights reserved.

引用

页码：1804 / 1815

页数：11

共 23 条

[1]

MARQUEZ FP., Photovoltaic plant condition monitoring using thermal images analysis by convolutional neural net⁃ work-based structure ［J］, Renewable Energy, 153, pp. 334-348, (2020)

[2]

CHEN H Y，, ZHOU Z., BAF-detector：an efficient CNN-based detector for photovoltaic cell defect detection［J］, IEEE Transactions on Industri⁃ al Electronics, 69, 3, pp. 3161-3171, (2022)

[3]

Chen H, Et al., Structure-aware-based crack defect detection for multicrystalline solar cells［J］, Measurement, 151, (2020)

[4]

TAO Z Y, YU Z J, LIN S., Research on crack de⁃ fect detection of solar cell based on PSO_SVM［J］, Journal of Electronic Measurement and Instrumenta⁃ tion, 35, 1, pp. 18-25, (2021)

[5]

GONZALEZ C, Et al., Automatic solar cell diagnosis and treat⁃ ment［J］, Journal of Intelligent Manufacturing, 32, 4, pp. 1163-1172, (2021)

[6]

AKRAM MW, JIN Y，, Et al., CNN based au⁃ tomatic detection of photovoltaic cell defects in elec⁃ troluminescence images［J］, Energy, 189, (2019)

[7]

CHEN H Y，, CHEN P，, Et al., Deep learn⁃ ing-based solar-cell manufacturing defect detection with complementary attention network［J］, IEEE Transactions on Industrial Informatics, 17, 6, pp. 4084-4095, (2021)

[8]

Detection of surface defects on solar cells by fusing Multi-channel convolution neural networks［J］, In⁃ frared Physics & Technology, 108, (2020)

[9]

KULKARNI D V., Anomaly de⁃ tection and segmentation for wafer defect patterns us⁃ ing deep convolutional encoder-decoder neural net⁃ work architectures in semiconductor manufacturing ［J］, IEEE Transactions on Semiconductor Manu⁃ facturing, 32, 2, pp. 250-256, (2019)

[10]

YANG W., Event-Based High Frame-Rate Video Reconstruction with A Novel Cycle-Event Network［C］, 2020 IEEE Internation⁃ al Conference on Image Processing（ICIP）, pp. 86-90

← 1 2 3 →