Critical Technical Study On Image Processing Of Zero Insulator Based On Infrared Thermography

Since entering the21st century, our country’s power grid scale and voltage level havebeen expanded and improved. If there is zero resistance insulator in the transmission line,equivalently part of the insulation is short-circuited, it will increase the insulator flashoverprobability. It will be a serious threat to the security of the transmission network,and will leadto the local power grid collapse and paralysis. And it will severely affect the country’sindustry and agriculture and people’life. Therefore, accurate detection of zero insulatorbecomes very important. With the development of infrared technology, infrared thermalimaging technology is widely used in electrical equipment fault detection, and it is also animportent method for zero value insulator detection. This paper is focus on critical technicalstudy on image processing of infrared zero insulator image.A new denoising method is proposed in the paper according to the characteristics ofinsulator infrared image with impulse noise. Firstly, make use of the pulse coupled neuralnetwork (PCNN) to detect the location of the impulse noise pixels and keep the non-noisepixels stable. Then, according to the characteristics of the impulse noise, the window size ofthe filter is adaptively determined by calculating the noise intensity of the image. The pixelswith maximum and minimum gray value in filtering window are excluded, using thesimilarity of the left pixels to calculate the weights. And a new weighted filtering algorithm isused to filter noise pixels. The image processing results shows that this method has excellentnoise reduction performance, and have good signal noise ratio, and at the same time keep theintegrity of the image detail very well.In order to segment the insulator string with zero value as much as possible from thewhole image, we circularly light the denoised images and use the pulse coupled neuralnetwork to segment the image. We get ignition images, and calculate the intra-class andinter-class variance, then stop igniting when get the biggest ratio. At this time, the number ofignition is the best number of the iterations, and we can get the best results of imagesegmentation. The experiments show that, we can use this method to completely andeffectively segment out the normal insulator infrared image and zero insulator infrared image,and the segmentation effect is obvious and with good performance. It provided a guarantee forthe subsequent feature extraction and accurate identification of the zero insulator.In summary, this paper solved the key technical problems about image denoising andimage segmentation of the insulator infrared image, improved the performance of imagedenoising and segmentation, made a good focudation for the subsequent identification of thezero value of insulator.