基于无人机的风机叶片表面缺陷自动检测方法
2024,32(11):72-79
摘要：风机叶片是风力发电系统的核心部件,受到气候条件、工作负荷等因素的影响,容易出现各类缺陷,如裂纹、磨损、腐蚀等。如果不能及时发现和解决这些缺陷,将导致风机性能下降、损坏甚至引发事故。为此,研究一种基于无人机的风机叶片表面缺陷自动检测方法。利用无人机搭载摄像机,飞到高空当中,拍摄空中运行的叶片图像。对叶片图像实施灰度化、去噪以及照度均衡化处理,提升图像质量。提取叶片图像中的几何特征和纹理特征。利用差异演化算法改进概率神经网络平滑参数,以优化后的概率神经网络为基础构建分类识别模型,将几何特征和纹理特征作为输入,计算每种类别的输出概率,将最大值响应原则将概率数值最大的类别作为判定的缺陷类别,以此实现风机叶片表面缺陷自动检测。结果表明:所研究技术应用下,杰卡德系数相对更高,说明该方法的检测结果更为准确；所花费时间相对更少,说明该方法的检测效率更高,可以更快地完成检测任务。
关键词：无人机;风机叶片;特征提取;改进概率神经网络;缺陷自动检测技术
Automatic detection method for surface defects of fan blades based on drones
Abstract：Wind turbine blades are the core components of wind power generation systems, which are susceptible to various defects such as cracks, wear, and corrosion due to factors such as climate conditions and working loads. If these defects cannot be detected and resolved in a timely manner, it will lead to a decrease in fan performance, damage, and even accidents. To this end, a method for automatic detection of surface defects on fan blades based on drones is studied. Using a drone equipped with a camera, fly high into the air to capture images of the blades moving in the air. Implement grayscale, denoising, and illumination equalization processing on leaf images to improve image quality. Extract geometric and texture features from leaf images. Using differential evolution algorithm to improve the smoothing parameters of probability neural network, constructing a classification recognition model based on the optimized probability neural network, taking geometric and texture features as inputs, calculating the output probability of each category, and using the maximum response principle to determine the defect category with the highest probability value, in order to achieve automatic detection of surface defects on wind turbine blades. The results show that under the application of the studied technology, the Jaccard coefficient is relatively higher, indicating that the detection results of this method are more accurate; The relatively less time spent indicates that this method has higher detection efficiency and can complete detection tasks faster.
Key words：Drones; Fan blades; Feature extraction; Improving probabilistic neural networks; Automatic defect detection technology
收稿日期：2024-02-19
基金项目：
