基于高光谱成像技术的绝缘子污秽度预测

高光谱成像技术能对绝缘子进行非接触式成像，且具有多波段、图谱合一等特点. 为此，本文提出一种基于高光谱成像技术的绝缘子污秽度预测方法. 首先，利用高光谱成像仪对绝缘子进行成像，得到400～1 000 nm波段范围内的高光谱图像数据，并进行黑白校正；然后，获取感兴趣区域（region of interest，ROI）的反射率光谱曲线，进行Savitzky-Golay平滑、对数或一阶导数变换的预处理. 最后，联合部分的真实样本标签数据分别建立基于支持向量机的绝缘子污秽度预测（support vector machines-insulator contamination degree prediction，SVM-ICDP）和基于偏最小二乘回归的绝缘子污秽度预测（partial least squares regression-insulator contamination degree prediction，PLSR-ICDP）模型. 从实验结果中可知，当预处理方法采用一阶导数变换时，所建立的绝缘子污秽度预测模型效果最佳，即SVM-ICDP模型准确率达到91.84%；PLSR-ICDP模型的均方根误差（root mean square error，RMSE）为0.024 1. 

Contamination Degree Prediction of Insulators Based on Hyperspectral Imaging Technology

Insulator image can be acquired by hyperspectral imaging technology in a non-contact way, and hyperspectral image has some advantages such as the properties of multi-band, and merging image and spectrum. For this reason, the paper proposes a method to predict contamination degree of insulators based on hyperspectral imaging technology. Firstly, the hyperspectral image in a band range of 400–1 000 nm is acquired by hyperspectral imaging system, followed by the black-and-white correction. Then, some reflectivity spectrum curves of region of interest (ROI) are extracted and further pre-processed by the methods such as the Savitzky-Golay smoothness, logarithm, or first derivative transformations. Finally, some labeled data of real samples are utilized to build support vector machines based insulator contamination degree prediction (SVM-ICDP) model and partial least squares regression based insulator contamination degree prediction (PLSR-ICDP) model, respectively. The experimental results show that when the first derivative transformation is selected as the pre-processing method, the performance of the ICDP model is superior to those of the others. More specifically, the accuracy of SVM-ICDP reaches 91.84%, and the root mean square error (RMSE) of PLSR-ICDP is 0.024 1.