基于迁移学习的光伏组件鸟粪覆盖检测  

BIRD DROPPINGS COVERAGE DETECTION OF PHOTOVOLTAIC MODULE BASED ON TRANSFER LEARNING

作　　者：李琼 吴文宝 刘斌[1] 刘君[1] Li Qiong;Wu Wenbao;Liu Bin;Liu Jun(School of Information Engineering,Nanchang Hangkong University,Nanchang 330063,China;Jiangxi Electric Power Construction Co.,Ltd.of China Power Construction Group,Nanchang 330001,China)

机构地区：[1]南昌航空大学信息工程学院,南昌330063 [2]中国电建集团江西省电力建设有限公司,南昌330001

出　　处：《太阳能学报》2022年第2期233-237,共5页Acta Energiae Solaris Sinica

基　　金：江西省科技厅重点研发项目(20192BBE50057)。

摘　　要：该文基于无人机光伏组件可见光图像采集,提出一种基于迁移学习的光伏组件鸟粪检测方法。方法首先基于掩膜区域卷积神经网络(Mask-RCNN)对光伏组件边界进行框选,再利用迁移学习策略,构建光伏组件鸟粪缺陷检测模型,实现鸟粪智能检测。利用上述方法,实现光伏组件鸟粪覆盖检测准确率为96.75%。In this paper,based on the visible light image acquisition of photovoltaic(PV)panel by using unmanned aerial vehicle,a bird droppings detection method of photovoltaic panel based on transfer learning is proposed. Firstly,the boundary of PV panel is selected based on Mask-RCNN,and then the bird droppings defect detection model of PV panel was constructed by using transfer learning strategy to realize intelligent detection of bird droppings. Using the above method,the detection accuracy of bird droppings coverage of PV panel is 96.75%.

关 键 词：迁移学习 光伏组件 掩膜区域卷积神经网络 缺陷检测 

分 类 号：TK51[动力工程及工程热物理—热能工程]