基于YOLOv5模型改进的太阳能电池片小尺寸缺陷精准检测算法  

Improved precise detection algorithm for small size defects in solar cell based on YOLOv5 model

作  者:张帅 高德东 赵德胜 孟祥纯 ZHANG Shuai;GAO Dedong;ZHAO Desheng;MENG Xiangchun(College of Mechanical Engineering,Qinghai University,Xining 810016,China)

机构地区:[1]青海大学机械工程学院,西宁810016

出  处:《青海大学学报》2025年第1期8-14,共7页Journal of Qinghai University

基  金:青海省光伏发电并网技术重点实验室2024年开放性课题(SGTYHT/23-JS-004)。

摘  要:在电致发光环境中检测太阳能电池片裂纹、碎片和功能失效等小尺寸和不规则故障的难度较大、精度较低,为此提出一种结合可变形卷积网络(DCNv2)和最小点距离交并比(MPDIoU)损失函数的改进YOLOv5模型。通过使用DCNv2替换原有的标准卷积C3模块,引入MPDIoU损失函数替换原有的CIoU,提高模型处理小尺寸复杂形状缺陷时的适应性和检测精度,减小预测框与实际框之间的定位误差。结果表明:改进YOLOv5模型的mAP@50的值为78.4%,较原始模型提高5.3个百分点;mAP@50-95的值为72.2%,较原始模型提高5.4个百分点,且模型的检测速度维持在每秒53帧。由此证明,以上改进不仅能提升模型在复杂背景中的鲁棒性和适应性,还能增强模型在检测小尺寸缺陷时的准确性和实时性。In the electroluminescent environment,detecting small size and irregular faults such as cracks,debris and functional failures of solar cells is difficult with low accuracy.This article proposes an improved YOLOv5 model that combines deformable convolutional network(DCNv2)and minimum point distance intersection over union(MPDIoU)loss function.By replacing the original standard convolution C3 module with DCNv2 and introducing MPDIoU loss function to replace the original CIoU,the adaptability and accuracy of the model in handling small size and complex defects were improved.The positioning error between the predicted boxes and the actual boxes was reduced.The results are as follows:The value of mAP@50 of the improved YOLOv5 model reached 78.4%,increased by 5.3 percentage points compared to the original model;the value of mAP@50-95 was 72.2%,5.4 percentage points higher than the original model,and the detection speed of the model was maintained at 53 frames per second.This proves that the above improvements can not only enhance the robustness and adaptability of the model in complex backgrounds,but also improve the accuracy and real-time performance of the model in detecting small size defects.

关 键 词:YOLOv5 电致发光 太阳能电池 缺陷检测 可变形卷积网络 最小点距离交并比 

分 类 号:TM914.4[电气工程—电力电子与电力传动]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象