Intelligent Identification Of Transmission Line Defects Based On FPGA

Transmission lines are an important infrastructure of the national power grid system,and their safe and stable operation is an important guarantee for power transmission.The state must spend a lot of manpower and materials every year on the inspection of transmission lines in order to effectively eliminate hidden dangers.However,with the continuous development of the power industry,the power supply area and the total length of the line have also increased.Traditional manual line inspection can no longer meet the actual requirements.New line inspection technology is needed to intelligently identify the targets to be detected on the transmission line and determine whether Flawed.The rapid development of image processing and deep learning technologies has provided feasibility for intelligent identification of defects.Further upgrades of chips such as FPGAs provide a hardware platform for digital image processing and hardware acceleration of convolutional neural networks.In this paper,the visible light image of the transmission line is used as the data source,and the FPGA development board is used as the hardware platform to complete the identification of components on the transmission line,the segmentation of the target image,and the final defect detection.First,hardware acceleration is used to implement the convolutional neural network,which is used to identify the target to be detected in the image of the transmission line and provides a basis for the classification detection of defects.First,the network model was built and trained on the PC,and it was determined that the network can identify the target components.The parallel operation of the reconvolutional neural network was improved,and it was successfully transplanted to the FPGA to achieve hardware acceleration of the convolutional neural network.Finally,experiments verify the feasibility of convolutional neural network recognition on a hardware platform.Secondly,an image segmentation method using histogram equalization,threshold segmentation,morphological filtering,and Sobel edge detection fusion is adopted.The target in the image is successfully separated from the background to obtain a target contour image that is convenient for defect detection.Use the equalization of the histogram to complete the preprocessing of the image,solve the problem of image exposure,and enhance the contrast between the object to be detected and the background in the image;use an improved local adaptive threshold segmentation algorithm to solve the problem of image segmentation under uneven lighting The ideal problem is to successfully segment the target to be detected from the image;use morphological filtering to eliminate image background noise and interference from small objects,expand the highlight area of the target to be detected,and enhance the image effect;use Sobel edge detection to extract the outside of the target The contours provide a basis for defect detection.On the FPGA side,a top-down modular design was used to complete the transplantation of the above algorithms.The image segmentation on the hardware platform was successfully implemented,and a good image segmentation effect was obtained in the experiment.Finally,a defect detection algorithm based on insulator contour information is proposed for insulator defects,and it is successfully implemented on FPGA to achieve the purpose of defect detection.Firstly,the upper and lower edge curves of the insulator are extracted from the contour image of the target,and then the curve is smoothed using the local weighted regression algorithm(LOWESS)to obtain the peak value of the insulator sheet.Finally,the distance relationship between adjacent wave peaks Determine whether there is a defect in the insulator and complete the inspection of the insulator.