绝缘子污秽等级的高光谱特征优化识别技术研究

为解决传统污秽检测方法对输电线路绝缘子污闪防治的局限性,采用非接触式、高分辨率的高光谱成像技术对污秽在线检测技术进行研究。为有效提取反应污秽度的光谱特征并削弱冗余与干扰信息的影响,提出一种小波包能量谱特征优化的绝缘子污秽等级识别技术。首先,对不同污秽度的绝缘子样品的光谱图像进行背景分割,提取均匀覆污区像素点光谱均值曲线；其次,对不同图像的光强均匀度差异、环境噪声进行预处理,并通过导数变换提升不同污秽等级间的可区分性。再次,对预处理后的谱线进行小波能量谱特征提取。最后,基于所提特征建立基于支持向量机(support vector machines, SVM)的污秽等级识别模型。实验结果表明,相比于采用全波段数据或PCA特征数据作为输入,基于小波能量谱特征建立的支持向量机(SVM)污秽等级识别模型对样品识别准确率达到99.8%。#$NL关键词：关高光谱成像；绝缘子污秽等级；小波包能量谱；支持向量机#$NL中图分类号：TM933

Optimization and identification technology of hyperspectral spectral features of insulator pollution levels

In order to solve the limitation of traditional pollution detection methods on transmission line insulator pollution flashover prevention and control, the online pollution detection technology of non-contact and high resolution hyperspectral imaging is studied. In order to extract the spectral characteristics of reaction pollution degree effectively and reduce the influence of redundancy and interference information, a new insulator pollution level identification technology optimized by wavelet packet energy spectrum features was proposed. Firstly, background segmentation was performed on the spectral images of insulator samples with different pollution degrees to extract the mean spectral curves of pixel points in the uniform contaminated area. Secondly, the difference of light intensity uniformity and environmental noise of different images were preprocessed, and the differentiability between different pollution levels was improved by derivative transformation. Thirdly, wavelet energy spectrum features were extracted from the pretreated spectral lines. Finally, based on the proposed feature based on support vector machine (SVM) pollution level recognition model. The experimental results show that, compared with using full-band data or PCA feature data as input, the pollution level recognition model based on the wavelet energy spectrum features can achieve 99.8% accuracy of sample recognition.#$NLKeywords: hyperspectral imaging; insulator pollution rating; wavelet packet energy spectrum; support vector machine