自适应图像分割的光伏板积灰状态识别  

Photovoltaic panel dust accumulation state recognition based on adaptive image segmentation algorithm

作　　者：白云瑞 魏霞[1] 黄德启[1] 孟凡森 BAI Yunrui;WEI Xia;HUANG Deqi;MENG Fansen(School of Electrical Engineering,Xinjiang University,Urumqi 830017,China)

机构地区：[1]新疆大学电气工程学院,新疆乌鲁木齐830017

出　　处：《现代电子技术》2022年第22期43-47,共5页Modern Electronics Technique

基　　金：国家自然科学基金资助项目(51468062)。

摘　　要：针对光伏电站普遍存在的光伏阵列中太阳能光伏板积灰难识别以致于清理不及时的问题,文中提出一种基于自适应图像分割算法识别光伏板的积灰状态。根据光伏板灰尘图像的YCbCr颜色空间特性,利用自适应图像分割算法将灰尘与光伏板图像分离,进而在旋转统一不变LBP算法下提取图像纹理特征,最终实现太阳能光伏板积灰状态的识别。再分别对不同算子半径与周围像素点个数进行实验比较,并对该算法与深度学习算法在准确率与识别时间方面进行对比分析。实验结果表明:旋转统一不变LBP算法对图像特征提取的效果最好,算法对光伏板积灰图像识别准确率可达88.56%,识别速度较ResNet50网络快90%以上;且占用计算机内存小,能够满足电站运维实时性要求,提高发电效率,降低运维成本,具有较高的实用性价值。In allusion to problem that it is difficult to identify the dust accumulation on the solar photovoltaic panels in the photovoltaic array in the photovoltaic power station,which leads to untimely cleaning,the photovoltaic panel dust accumulation state identification based on adaptive image segmentation algorithm is proposed.According to the YCbCr color space characteristics of photovoltaic panel dust image,the adaptive image segmentation algorithm is used to separate dust from photovoltaic panel image,so as to extract the image texture features by means of the rotation unified invariant LBP(local binary pattern)algorithm.The dust accumulation state of solar photovoltaic panels is recognized.The comparison experiments of different operator radii and the number of surrounding pixels were carried out,and the comparison analysis of the accuracy and recognition time of the algorithm and the depth learning algorithm was conducted.The experimental results show that the rotation unified invariant LBP algorithm has the best effect in image feature extraction.The accuracy of the algorithm for identifying the image of dust accumulation on photovoltaic panels can reach 88.56%,its recognition speed is more than 90%faster than that of ResNet50 network,and it occupies less computer memory.It can meet the real-time requirements of power station operation and maintenance,improve power generation efficiency,reduce operation and maintenance costs,and has high practical value.

关 键 词：光伏电站 光伏板积灰 自适应图像分割 纹理特征提取 LBP算法 图像识别 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TP272[电子电信—信息与通信工程]