Abstract:The surface of naturally contaminated insulators is not completely covered by pollution and contains insulator material information. According to the detection principle of hyperspectral technology, the differences between porcelain, glass and silicone rubber insulator materials will be reflected in hyperspectral spectral line data, so it is hard to detect the contamination grades by the same model. Modeling the measured spectra of each material sample separately requires a large number of samples with known spectra and measured values, high modeling cost, long time and low efficiency. This paper realizes the rapid detection of the contamination grades of different material insulators through the piecewise direct standardization（PDS） model transfer method. Firstly, test samples with different materials and different contamination grades were prepared according to relevant standards, and their hyperspectral data were obtained by the hyperspectral data acquisition platform, and their salt density and ash density values were measured using the equivalent salt deposit density. Then, the samples hyperspectral data after removing the interference were obtained through preprocessing operation. Finally, the main model of contamination grades detection of support vector machine is established based on the data of glass insulator, and the PDS algorithm is used to transfer the data of porcelain and silicone rubber material samples. The contamination grades detection method of different material insulation samples are obtained. Through spectral line analysis and test verification, the following conclusions are drawn:（1） Under the same contamination grades, the hyperspectral lines of different material insulators have obvious differences, which are shown in the positions and changing trends of absorption peak and reflection peak. In the case of different contamination grades of the same material, the difference of spectral lines is mainly reflected in the amplitude. The higher the contamination grades, the lower the spectral line amplitude.（2） The support vector machine（SVM） contamination grades detection model was established by the data of glass insulation samples, and the detection accuracy of the contamination grades of glass samples was 98.3%. PDS algorithm can effectively reduce the difference between the spectral lines of porcelain, silicone rubber and glass samples. After the model transfer, the accuracy of the model in classifying the contamination grades of porcelain samples is increased from 25.0% to 95.0%, and the accuracy of the model in classifying the contamination grades of silicone rubber samples is increased from 25.0% to 91.7%.（3） The detection method proposed in this paper can detect the contamination grades of artificial contaminated insulators of different materials, and the accuracy is 83.3%.

A Contactless Insulator Contamination Levels Detecting Method Based on Infrared Images Features and RBFNN.

Insulator Contamination Perception Based on Feature Fusion of Infrared Image and Meteorological Parameters

A residual convolutional neural network with multi-block for appliance recognition in non-intrusive load identification

An Insulator in Transmission Lines Recognition and Fault Detection Model Based on Improved Faster RCNN

Two-stage Convolutional Neural Network for Joint Removal of Sensor Noise and Background Interference in Lock-in Thermography

An Improved Faster R-CNN for Devices Detection in Railway 4C System

Catenary insulator defect detection based on contour features and gray similarity matching

Railway insulator defect detection with deep convolutional neural networks

Contamination Grades Detection Method of Ceramic Insulator Based on Hyperspectral Technology

A Multi-feature Fusion-based Deep Learning for Insulator Image Identification and Fault Detection

Detection Method of Contamination Grades of Insulators with Different Materials Based on Hyperspectral Technique

Arbitrary-Oriented Detection of Insulators in Thermal Imagery via Rotation Region Network

Automatic Fault Diagnosis of Infrared Insulator Images Based on Image Instance Segmentation and Temperature Analysis

Monitoring of Composite Insulators in Transmission Lines: A Hydrophobicity Diagnostic Method Using Aerial Images and Residual Neural Networks

An Extreme-Learning-Machine-Based Hyperspectral Detection Method of Insulator Pollution Degree.

Image Detection of Insulator Defects Based on Morphological Processing and Deep Learning

Insulators Recognition Based on Enhancement of High-Level Feature with Improved Attention Method

Detail R-CNN: Insulator Detection Based on Detail Feature Enhancement and Metric Learning

Insulator Defect Recognition Based on Global Detection and Local Segmentation

Method for Detecting the Pollution Degree of Naturally Contaminated Insulator Based on Hyperspectral Characteristics