Research On Diameter Measurement And Defect Detection Method Of Transmission Line Based On Machine Vision

In the line maintenance and theoretical calculation of power line loss in medium and low voltage distribution grid,it is often necessary to determine the diameter parameters of the wires.However the increase of line ledgers will easily lead to insufficient information update,and operational workers cannot obtain it in time.The wires in operation cannot be directly contacted to measure their diameters for inspection,which is inconvenient to inspection and maintenance work.The steel-cored aluminum stranded wires are mostly used in transmission and distribution grid overhead powerlines in our country,and the operating environment is complex.The surface of the wires can be easily oxidized and corroded,which lead to problems such as loose strands and broken strands caused by the decline of mechanical strength.It is an important task of line inspection to measure the diameter of the wire accurately,and to detect and eliminate potential safety hazards in time.In recent years,the inspection method that using UAVs to capture powerline images and processing images through machine vision technology improves efficiency,and enables non-contact measurement of wire diameter and automatic detection of wire defects.In this paper the basic knowledge of digital image preprocessing,morphological processing and image segmentation was firstly introduced.Then comparing the difference of the extraction effect of the wire under the condition of front light and back light by threshold segmentation.And accurate segmentation and extraction of power transmission wire in visible light images with complex environment was achieved based on the improved weighted chromatic aberration method,combined with Otsu threshold segmentation and image dilation and erosion.Afterwards,based on the visual measurement technology and the imaging model of the camera,a non-contact diameter measurement method of the transmission wire is proposed.Based on the extraction of the wire in the image,the straight line equation of the wire edge and the wire width in the image are obtained by edge detection and line fitting.Then combined with the camera focal length,object distance and other parameters,the true diameter of the wire is calculated.The mothed is verified through experiments,and the cause of the error is analyzed and be corrected.The measurement error of the wire diameter after correction is within 1mm.Finally,A method for wire defect detection using an improved sliding window is proposed.A convolutional neural network was constructed based on the residual module.The trajectory and size of the sliding window was calculated through the edge straight line equation of the wire,realizing the cropping of the sub-image of the transmission wire,and uses them as a training dataset to train the neural network.The network can identify three wire state: normal,corrosive and loose strands,whose accuracy rate reaches 98.12%.So the automatic detection of wire defects is realized.