Research On Surface Roughness Classification Recognition Of Roof Insulator Based On Hyperspectral Imaging Technology

With the continuous improvement of high-speed train running speed and the expansion of high-speed railway network coverage area,China’s railway transport capacity has been greatly improved.High-speed railway plays an important supporting role for China’s economy and it is of great significance to ensure the safe and stable operation of high-speed trains.The high-speed railway,represented by Lanzhou-Xinjiang railway line,passes through the sandstorm areas like gobi and desert.When the high-speed trains run in the strong wind and sand area,the roof insulators as important high-voltage insulation components will be impacted by high-speed dust particles,causing abrasion damage of the roof insulators and increasing their surface roughness.The increase of surface roughness of the roof insulators will cause changes and deterioration of water repellency,fouling characteristics and insulation performance,which may cause hidden dangers to the safe and stable operation of the trains.After the train returns to the factory for cleaning,workers mainly rely on observation to judge the surface condition of roof insulators.There is no reliable on-line detection method to accurately grasp the surface roughness of roof insulators.Therefore,it is necessary to propose a non-contact insulator surface roughness detection method which can be used online to help workers grasp the surface roughness of roof insulators and ensure the safe and stable operation of trains.In this paper,the surface roughness of roof insulators is classified and identified based on hyperspectral imaging technology.This paper analyzes the traditional surface roughness optical detection methods,then combines the principle of high spectral imaging technology,and explores the theoretical basis of high spectral imaging technology for the surface roughness detection of the roof insulators.Then the relationship between the surface roughness of roof insulators and their hyperspectral image data is studied,which provides theoretical support for the establishment of a method for measuring the surface roughness of roof insulators based on hyperspectral imaging technology.A hyperspectral image detection platform is set up,and the test samples are prepared.Then the traditional surface roughness detection method is used to measure the prepared test samples.The surface roughness grade calibration is completed with reference to the national standard,and the original hyperspectral image data of the test samples are extracted.The original hyperspectral image data are pre-processed by black-and-white correction,multivariate scattering correction and S-G smoothing filter respectively.The dimension reduction of hyperspectral image data is completed by continuous projection algorithm.The processing and optimization methods of hyperspectral image data are studied.The classification accuracy of data modeling for different supervised classification methods,different pre-processing methods and combinations of different characteristic bands has been compared respectively,and the combination of classification modeling method and data processing method under the optimal condition has been obtained.Then we established a discriminant model for the surface roughness level of roof insulators,completed parameter optimization to further optimize the model,and used test sample data for verification,and completed a grading detection method for the surface roughness of roof insulators based on high spectral imaging technology.This method can be used for on-line inspection of the cleaned insulators and help workers to grasp the surface roughness of the insulators on the roof of high-speed trains in time.