Research On Transmission Line Detection Based On Machine Vision

At present,artificial intelligence has been widely used in the power industry,and transmission line inspection based on machine vision is one of the typical applications.The image of transmission line is collected by UAV,and then the image is analyzed and processed by artificial intelligence method to distinguish line anomalies.In the system,the identification and fault detection of transmission equipment is one of the key technologies.The image detection targets of transmission line patrol inspection mainly include insulator,damper and spacer.For the identification of transmission equipment,two types of transmission equipment identification algorithms based on ensemble learning and depth learning are proposed.A defect detection algorithm based on morphology is proposed for insulator self-explosion defect detection.In the identification algorithm based on ensemble learning,the proposed candidate box generation algorithm is used to obtain the candidate boxes of transmission equipment;A classifier based on ensemble learning is designed to detect the images captured by candidate frames;A redundant prediction frame elimination algorithm is proposed to remove the redundant prediction frames generated in the detection process;Realize the identification of transmission equipment.Experiments show that the proposed redundant prediction frame elimination algorithm can better screen the localization frame,and the detection accuracy is improved by 0.03.The proposed recognition algorithm based on ensemble learning has more advantages than other detection methods based on shallow classifiers.In the transmission equipment algorithm based on deep learning,Faster RCNN is used as the basic framework,and the improved residual network is used as its feature extraction network.A clustering algorithm is proposed to analyze the labeled boxes of samples to obtain the preset anchor size in the network.Experiments show that the clustering algorithm can improve the m AP value of the model for transmission equipment detection by 0.83%.The proposed deep learning algorithm has more powerful power transmission equipment identification ability by comparison.Based on the identification of transmission equipment,the self-explosion defects of insulators are detected,and an insulator defect detection algorithm based on morphology is proposed.Firstly,an image segmentation algorithm is proposed to extract insulators from the target image.Then the spatial shape adjustment algorithm of insulator is proposed to normalize the shape of insulator and simplify the detection process.Finally,a mathematical model is established to describe the state of the insulator and realize the defect detection of the insulator.This algorithm can not only locate the defect location of insulator effectively,but also detect the damage degree of insulator.Using the data set collected on site,the two types of transmission equipment identification algorithms based on integrated learning and depth learning and the insulator defect detection algorithm based on morphology are tested and verified.The test results show that the three new algorithms have certain advantages over other algorithms,and can meet the needs of the line inspection system.