基于Y0L0v5的旋转边界框电容器目标检测

Object Detection of Rotated Bounding Boxes for Capacitors Detection Based on YOLOv5

打开文本图片集

,， 313119

中图分类号：TP391.4 文献标识码：A 文章编号：2096-4706（2025）11-0033-05

Object Detection of Rotated Bounding Boxes for Capacitors Detection Based on YOLOv5

ZHANG Zhihao1, YANG Xuejun1, SHEN Mouquan', HU Jiwei², KE Yun², LI Chaochao² (1.CollegeofElectricalEngineeringandControlScience,NanjingTech University,Nanjing211816,China; 2.ChangxingHuaqiangElectronicsCo.,Ltd.,Huzhou 313119, China;)

Abstract:To solve the problem that classicalYOLOv5object detection algorithm can only achieve object localization with horizontalrectangular bounding boxes,thispaper designsanobjectdetection methodof rotated bounding boxes for capacitors based on the YOLOv5s model.This method transforms the angle prediction problem from aregresion problem to aclassificationproblem by using circular smooth labels,and describes the lossfunction ofangle prediction using binary cross-entropyloss.Additionally,theorginal traningdataisexpandedtroughreplication,rotationtransformation,and stitchingtoimprovetheaccuracyand generalizationabilityofthemodel.Experimentalresultsonarealcapacitorsdataset showthat the improved object detection algorithm ofrotated bounding boxes forcapacitors achieves an average accuracy of 83.5% .Compared withthe original YOLOv5 model,the predicted object bounding boxes are more consistent with the actual rectangular contours of the capacitors.

Keywords: Object Detection; rotated bounding box;Deep Learning; capacitor

0 引言

在环境污染和石油危机双重压力下，能源结构转型和清洁可再生能源发电技术成为能源发展的重要方向，电能存储逐渐成为智能电网和构建能源互联网的关键技术[1]，电容器的地位日益突出。（剩余10598字）