摘要
为提高电站光伏组件缺陷识别的精度,提出一种基于无人机图像特征学习算法的缺陷识别技术。对此,以SSD网络为基础框架,引入深度残差结构和使用三分支特征融合替代双分支特征融合的方法改进SSD网络;然后利用改进的SSD网络对电站光伏组件缺陷进行识别。仿真结果表明,本方法提高了电站光伏组件裂纹、划痕、缺角等不同缺陷的识别精度,平均识别精度达到97.11%,且具有较快的识别速度,平均识别处理时间达到30.22帧/s;识别网络能对大、中、小不同尺度的电站光伏组件缺陷图像进行识别。
To improve the accuracy of defect recognition of photovoltaic modules in power plants,a defect recogni⁃tion technology based on UAV image feature learning algorithm is proposed.In this regard,based on the SSD net⁃work framework,the deep residual structure and the method of using three-branch feature fusion instead oftwo-branch feature fusion are introduced to improve the SSD network.Then the improved SSD network is used toidentify the defects of photovoltaic modules in power plants.The simulation results show that this method improvesthe recognition accuracy of different defects such as cracks,scratches and missing angles of photovoltaic modules inpower plants.The average recognition accuracy reaches 97.11%,and has a faster recognition speed.The averagerecognition processing time reaches 30.22 frames/s;the recognition network can identify the defect images of photo⁃voltaic modules in power plants with large,medium and small scales.
作者
岳攀
熊开智
李志飞
周家麒
陈继发
洪流
YUE Pan;XIONG Kaizhi;LI Zhifei;ZHOU Jiaqi;CHEN Jifa;HONG Liu(Yalong River Hydropower Development Co.,Ltd.,Chengdu 610051,China;SNEGRID Technology Co.,Ltd.,Hefei 230088,China)
出处
《粘接》
2025年第10期171-174,共4页
Adhesion
关键词
光伏组件
缺陷识别
无人机图像
SSD网络
photovoltaic modules
defect identification
drone images
SSD network
