Research On Image Recognition Algorithm For Broken Strands And Damage Faults Of Transmission Conductor

The conductor of transmission line is a major component of transmission line which is the artery of power systems.The broken strands and damage faults on transmission conductor,caused by many elements such as external force,meteorological disaster and even wiring aging,could bring serious threats to the safe operation of the power system.The running status of the transmission conductor could be detected by the real-time inspection system based on the videos or images obtained from aerial inspection of transmission lines,to effectively guarantee the safe operation of the power system.There for,the research of the method of the detected of the broken strands and damage faults of conductor based on Unmanned Aerial Vehicle(UAV)inspection has a very important practical significance.The inspection task consists of two parts as the image preprocessing and the fault identification of the transmission conductor.(1)Image preprocessing.Due to the different image features of the conductor area and the background,one new color model conversion method based on the gray variance normalization is proposed in the paper which could effectively highlight the conductor region in the image.Subsequently,comparing with the traditional image segmentation algorithm,an improved K-means segmentation algorithm combined with morphological processing is adopted to achieve the accurate extraction of the conductor region.(2)Fault identification.1)The detected of the damage faults conductor.Firstly,Due to the degree of dispersion that the grayscale value of the damage regions of the conductor surface is more prominent than the other areas of the image,the longitudinal suspect positons of defects are first located by analyzing the projection profile of the normalized results of the grayscale variance of each transversal line.Each longitudinal suspect interval determines a subimage,and then,transversal suspect positions are obtained based on the transversal profile in the subimage.At last,each pair of longitudinal and transversal intervals forms a suspect rectangle,and the rectangle of the damage conductor is recognized by a thresholding rule.2)The detected of the broken strands conductor.Firstly,according to the cycle characteristics of the conductor surface,the Z-shaped waveform which could represent a single conductor is obtained by the denoising and threshold processing of the axes which consists of the center axis of the conductor and its upper and lower symmetry axes.Then,the suspect broken strands positons of each axes are extracted by comparing with the actual width of the single wire.Finally,the number of the broken strands of the transmission conductor is recognized by the maximum value of each suspect breakage positons and the actual width of the individual conductor.The experimental results demonstrate that,the new color model transformation method proposed in the paper is able to effectively highlight the conductor area in the image,as well lay a good foundation for the detection of the broken strands and damage of the transmission conductor.In 95.72% of image,the proposed method of the detection of the damage faults of transmission conductor worked well,and the proposed algorithm of the detection of the broken strands faults which simplifies the problem of the detection of the broken strands faults into the problem of the processing of one-dimensional data could detect the broken strands faults of transmission conductor accurately and rapidly in 93.60% of image.