Research On Insulator Defect Detection Method Based On Image Fusion

With the advent of the era of smart grid,the traditional manual line inspection method is time-consuming,laborious and dangerous,which has been gradually replaced by automatic detection technology.Aiming at the insulator fault detection in transmission line,image fusion technology is used.The infrared image and visible image of insulator are obtained by different sensors,and different information is fused by image fusion technology,and the insulator defects are analyzed according to the fused image.The main contents of this paper are as follows.Firstly,the current situation of insulator detection is analyzed,and the image fusion method is proposed for insulator defect detection.Image fusion theory is studied systematically,including image fusion levels,methods and evaluation criteria.This paper focuses on the four methods of pixel level image fusion,which are spatial domain,multi-scale transformation,sparse representation and neural network.Each method has its own advantages and disadvantages.The multi-scale transformation method has better fusion effect,so it is proposed to use multi-scale transformation method for image fusion.Secondly,we study the non subsampled contour transform(NSCT)and the non subsampled shearlet transform,NSCT is composed of two parts: the non lower sampling pyramid filter bank and the directional filter bank.The NSST is composed of two parts: the non lower sampling pyramid filter bank and the shear wave filter bank.The fusion rule of sub-band image is studied,and the FT algorithm is improved by using two-dimensional modal decomposition.The salient features of the image are extracted by the improved ft algorithm,and the local region matching degree of the low-frequency image is calculated by using the salient features of the image.If the matching degree is high,the weighted average method is used for fusion,if the matching degree is low,the gray value maximization method is used for fusion,and the spatial frequency method is used As the input of pulse coupled neural network(PCNN),Laplace energy and the link strength of PCNN,the improved PCNN is used to calculate the ignition frequency of high-frequency image pixels,and the gray value of fusion image is determined according to the ignition frequency.The infrared image and visible image of defective insulator are collected and processed,and the parameters in the image fusion process are set.Four fusion methods are used to carry out comparative experiments,including weighted average method,wavelet transform method,NSCT transform method and NSST transform method.The evaluation standard is used to evaluate the fusion results and the calculation speed is faster.Finally,according to the characteristics of fusion image,the types of defects in insulator are judged and the accuracy of different defect types detection is analyzed.