摘要
为实现对风电机组叶片表面缺陷检测的智能化,该研究应用无人机技术、图像视觉技术和深度学习算法,建立风电机组叶片缺陷检测系统,提高了对叶片上开裂缺陷的检测精度;系统使用sobel算子计算图像横向和纵向的梯度,并对图像进行阈值分割和去噪处理;构建深度学习模型提取图像缺陷的特征信息,加入了SPP-Net网络进行卷积操作,增加了模型的输入数据尺度,得到特征图后在利用PRN网络筛选特征图;实验结果显示该研究系统能够去除大量无用的背景信息,开裂缺陷部位的特征信息保留完整,对验证集中的图像进行测试后,该研究系统识别出的开裂缺陷数最高可达到50个。
In order to realize the intelligent detection of the surface defects of wind turbine blades,the research technologies of unmanned aerial vehicle(UAV),the image vision and deep learning algorithms are applied to establish the wind turbine blade defect detection system,which improves the detection accuracy of cracking defects on the blades.The system uses the sobel operators to calculate the horizontal and vertical gradients of the image,and performs threshold segmentation and denoising processing on the image.This paper constructs a deep learning model to extract the feature information of image defects,joins the SPP-Net network for convolution operation,increases the input data scale of the model,and uses the PRN network to filter the feature map after obtaining the feature map.The experimental results show that the research system can remove a large amount of the useless background information,and the characteristic information of the cracking defects is completely kept.After testing the images in the verification set,the number of cracking defects identified by the research system can reach up to 50.
作者
董礼
韩则胤
王宁
王恩路
苏宝定
DONG Li;HAN Zeyin;WANG Ning;WANG Enlu;SU Baoding(CGN New Energy Holdings Co.,Ltd.,Beijing 100000,China)
出处
《计算机测量与控制》
2022年第8期142-146,154,共6页
Computer Measurement &Control
