改进SSD算法的光伏板污损检测  

Photovoltaic panel contamination detection based on improved SSD algorithm

作　　者：汪会达 张栋(指导)[1] 肖焰辉 张潇云 刘宇 陈德明 WANG Huida;ZHANG Dong;XIAO Yanhui;ZHANG Xiaoyun;LIU Yu;CHEN Deming(School of Mechanical Engineering,Shanghai Dianji University,Shanghai 201306,China)

机构地区：[1]上海电机学院机械学院,上海201306

出　　处：《上海电机学院学报》2025年第1期52-57,共6页Journal of Shanghai Dianji University

摘　　要：针对无人机搭载嵌入式设备的算力有限以及现有光伏板缺陷检测模型参数量大、计算复杂的问题,提出一种基于深度学习的轻量化检测方法。该方法以SSD算法为基础,采用Mobile NetV3作为主干网络,旨在减少模型的参数量和计算复杂度。同时引入一种注意力机制,以增强模型的特征提取能力,提高检测的准确率。此外,对自制数据集进行数据增强以丰富数据集的内容,从而增强模型的泛化能力。结果表明:相较于原SSD算法,改进后的算法在计算量上减少了68.9%,平均精度提升了4.3%,检测速度达到了45.6 FPS。改进的算法能够快速准确地检测出光伏板表面的污损,有助于提高光伏电站的发电效率。The visible light inspection of photovoltaic power stations is performed using unmanned aerial vehicles(UAVs).To address the challenges of limited computational power in embedded devices mounted on UAVs and the large parameter size and computational complexity of existing defect detection models for photovoltaic panels,a lightweight detection method based on deep learning is proposed.This method adopts the SSD algorithm with MobileNetV3 as the backbone network,aiming to reduce the model's parameter size and computational complexity.Additionally,an attention mechanism is introduced to enhance the model's feature extraction capability and improve detection accuracy.Moreover,a self-constructed dataset is augmented to enrich its content,thereby improving the model's generalization ability.The results demonstrate that,compared to the original SSD algorithm,the improved algorithm reduces computational cost by 68.9%,increases average precision by 4.3%,and achieves a detection speed of 45.6 fps.The proposed algorithm can quickly and accurately detect surface contamination on photovoltaic panels,contributing to improved power generation efficiency in photovoltaic power stations.

关 键 词：目标检测 光伏板 深度学习 轻量化 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]