基于CBS-YOLOv8的光伏组件缺陷检测

郭建; 王蕊; 钟琪峰

doi:10.19912/j.0254-0096.tynxb.2025-0730

English Version Journal Login

您当前的位置：

首页 >

文章列表页 >

基于CBS-YOLOv8的光伏组件缺陷检测

更新时间：2026-10-15

- 基于CBS-YOLOv8的光伏组件缺陷检测
- Vol. 47, Issue 9, Pages: 280-287(2026)
- 作者机构：
- 作者简介：
- 基金信息：
- DOI：10.19912/j.0254-0096.tynxb.2025-0730
  CLC： TP183
- Online First：10 October 2026，
  
  Published：2026
- 稿件说明：
移动端阅览
郭建, 王蕊, 钟琪峰. 基于CBS-YOLOv8的光伏组件缺陷检测[J]. 2026, 47(9): 280-287.
DOI：

郭建, 王蕊, 钟琪峰. 基于CBS-YOLOv8的光伏组件缺陷检测[J]. 2026, 47(9): 280-287. DOI： 10.19912/j.0254-0096.tynxb.2025-0730.

摘要

针对光伏组件缺陷检测中存在的检测精度低、小目标漏检率高及网络参数量偏大等问题

提出一种基于YOLOv8的轻量级改进模型CBS-YOLOv8（CBAM-BiFPN-SIoU-YOLOv8）。在特征提取阶段通过嵌入的CBAM注意力模块强化缺陷区域的特征表征

在多尺度特征融合中采用双向特征金字塔网络（BiFPN）

并改良SIoU损失函数提升边界框回归精度

进一步结合知识蒸馏技术构建教师-学生协同训练框架

实现模型压缩与性能迁移。实验结果表明：改进后模型在自建光伏组件EL图像数据集上的mAP0.5达到87.50%

较原始YOLOv8提升7.20个百分点;经蒸馏压缩的CBS-YOLOv8模型参数量降低75.42%

检测精度仅下降0.45个百分点

在保持检测精度的同时满足轻量化需求。

Abstract

Aiming at the problems existing in the defect detection of photovoltaic modules

such as low detection accuracy

high missed detection rate of small targets and large number of network parameters

a lightweight improved model CBS-YOLOv8（CBAM-BiFPN-SIoU-YOLOv8） based on YOLOv8 is proposed. In the feature extraction stage

the feature representation of the defect area is strengthened through the embedded CBAM attention module. In multi-scale feature fusion

the bidirectional feature pyramid network （BiFPN） is adopted

and the SIoU loss function is improved to enhance the bounding box regression accuracy. Furthermore

knowledge distillation technology is integrated to construct a teacher-student collaborative training framework to achieve model compression and performance transfer. The experimental results show that the mAP of the improved model on the self-built photovoltaic module EL image dataset reaches 87.50%

which is 7.20 percentage points higher than that of the original YOLOv8. The number of parameters of the CBS-YOLOv8 model after distillation and compression decreased by 75.42%

and the detection accuracy only dropped by 0.45percentage points

meeting the lightweight requirements while maintaining the detection accuracy.

关键词

Keywords

references

Zhu J H, Zhou D Q, Lu R S, et al.C2DEM-YOLO: improved YOLOv8 for defect detection of photovoltaic cell modules in electroluminescence image[J]. Nondestructive Testing and Evaluation, 2025, 40(1): 309-331.

霍爱清, 张书涵, 杨玉艳, 等. 密集交通场景中改进YOLOv3目标检测优化算法[J]. 计算机工程与科学, 2023, 45(5): 878-884.

田浩, 周强, 贺晨龙. 基于多尺度特征融合的光伏组件缺陷检测[J]. 计算机工程与应用, 2024, 60(3): 340-347.

Tan L, Liu Z K, Huang X K, et al.A transformer-based UAV instance segmentation model TF-YOLOv7[J]. Signal, Image and Video Processing, 2024, 18(4): 3299-3308.

孙铁强, 于洪健, 宋超, 等. 轻量级重参数化的PCB缺陷检测算法[J]. 组合机床与自动化加工技术, 2025(4): 118-123, 128.

杜娟, 南晓林, 晋美娟, 等. 改进YOLOv8的带钢表面缺陷检测技术[J]. 组合机床与自动化加工技术, 2025(4): 167-172.

张冀, 王文彬, 余洋. 基于RFCARep-YOLOv8n的光伏电池缺陷检测算法[J]. 计算机工程与应用, 2025, 61(3): 131-143.

刘义艳, 郝婷楠, 贺晨, 等. 基于DBBR-YOLO的光伏电池表面缺陷检测[J]. 图学学报, 2024, 45(5): 913-921.

Wang Y, Chen X Q, Li J Q, et al.Convolutional block attention module-multimodal feature-fusion action recognition: enabling miner unsafe action recognition[J]. Sensors, 2024, 24(14): 4557.

赵陆阳, 刘长良, 刘卫亮, 等. 基于CBAM-DSC-UNet模型的时空风速预测算法[J]. 太阳能学报, 2024, 45(10): 497-505.

朱永红, 吴松涛. 基于改进的YOLOv8n瓷砖缺陷检测与分类方法[J]. 陶瓷学报, 2024, 45(6): 1255-1264.

Lawal O M.Real-time cucurbit fruit detection in greenhouse using improved YOLO series algorithm[J]. Precision Agriculture, 2024, 25(1): 347-359.

廖丽瑛, 刘洪. 基于改进YOLOv8的绝缘子自爆缺陷检测[J]. 电子测量技术, 2024, 47(18): 138-144.

杨丽, 杨晨晨, 杨耿煌, 等. 改进YOLOv8的光伏电池缺陷检测算法[J]. 电子测量技术, 2025, 48(1): 92-99.

任喜伟, 余杰, 韩欣, 等. 基于改进YOLOv5s网络的光伏组件缺陷检测方法[J]. 太阳能学报, 2025, 46(3): 428-434.

Li X Y, Wang F P, Wang W, et al.DM-YOLOX aerial object detection method with intensive attention mechanism[J]. The Journal of Supercomputing, 2024, 80(9): 12790-12812.

张润梅, 潘晨飞, 陈梓华, 等. 改进YOLOv8的焊缝表面缺陷检测算法[J]. 光电工程, 2025, 52(3): 240296.

刘雄彪, 杨贤昭, 陈洋, 等. 基于CIoU改进边界框损失函数的目标检测方法[J]. 液晶与显示, 2023, 38(5): 656-665.

麻斯亮, 许勇. 最小点距离的边界框回归损失函数及其应用[J]. 小型微型计算机系统, 2024, 45(11): 2695-2701.

Xing Z Q, Chen X, Pang F Q.DD-YOLO: an object detection method combining knowledge distillation and differentiable architecture search[J]. IET Computer Vision, 2022, 16(5): 418-430.

Gou J P, Yu B S, Maybank S J, et al.Knowledge distillation: a survey[J]. International Journal of Computer Vision, 2021, 129(6): 1789-1819.

张琳, 薄敬东, 龚瑞昆, 等. 基于改进YOLOv5s的红外船舶目标检测算法[J]. 电光与控制, 2024, 31(9): 38-44.

罗志聪, 何陈涛, 陈登捷, 等. 基于轻量化YOLOv8s-GD的自然环境下百香果快速检测模型[J]. 农业机械学报, 2024, 55(8): 291-300.

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

基于HCF-YOLO的光伏组件红外图像热斑故障检测

基于改进YOLOv8的光伏缺陷快速检测

基于知识蒸馏的红外光伏组件缺陷检测模型

PHOTOVOLTAIC MODULE DEFECT DETECTION METHOD BASED ON IMPROVED YOLOV5S NETWORK

基于时空双流网络与多重注意力的光伏功率预测

Related Author

秦怡鸣

尹丽菊

高晓宁

王峰

赵永辉

李振

金帅

颜培钰

Related Institution

Shaanxi Joint Laboratory of Artificial Intelligence

Shaanxi University of Science and Technology, School of Electronic Information and Artificial Intelligence

Liaoning Solar Energy Research and Application Co., Ltd.

School of New Energy, Shenyang Institute of Engineering

Jiangxi Electric Power Construction Co., Ltd. of China Power Construction Group

AI问答

⁰