The differences in pollution components of transmission line insulator will affect the flashover voltage

so the on-line identification of insulator pollution components is of great significance to the prevention and control of pollution flashover. We proposed an identification method of insulator pollution components based on hyperspectral technology. Firstly

through the collection of hyperspectral images of polluted samples with different components

935～1 725 nm hyperspectral curves were extracted from label samples after black-and-white correction. Secondly

the characteristic wavelengths of different pollution components were extracted by competitive adaptive reweighting sampling algorithm. Finally

the random forests classification models

based on the characteristic spectrum of training data and the full-band training data

respectively

were established to realize the identification of pollution components of test samples. The results show that only the total reflectance of hyperspectral curves obviously changes with the same pollution components

while the total reflectance

peak and change trend obviously change with different pollution components

and the random forests classification model based on the characteristic spectrum of training data and the full-band training data can identify pollution components

and the identification accuracy of single pollution component reach 91.7% and 93.3%

respectively

while the identification accuracy of mixed pollution reach 90% and 75%

respectively. Therefore

this method can be applied to the fast

accurate and on-line identification of insulator pollution components

providing a technical reference for the on-line detection.