Wire strand-splitting identification method and system based on neural network

Abstract

The invention discloses a wire strand-splitting identification method and system based on a neural network. The method relates to the technical field of image recognition technology and machine learning, and comprises the following steps: establishing a defect sample system which comprises a wire strand-splitting fault picture and a wire fault picture detection result; generating a label file by utilizing the manually labeled fault picture, and converting the label file to generate a data training set and a test set; utilizing an SSD target detection algorithm and the fault picture annotationinformation to carry out classification regression on the candidate frames, and utilizing a MobileNet network and a data set to adjust a hyper-parameter training wire strand opening fault detection model; loading the wire strand-splitting fault detection model into wire strand-splitting defect identification intelligent equipment; and the wire strand-splitting defect identification intelligent device transmitting the detected wire strand-splitting picture to the server, so that the defect sample library is expanded.

Description

Technical field [0001] The invention relates to the field of image recognition technology and machine learning technology, and in particular to a method and system for identifying strand openings based on neural networks. Background technique [0002] With the continuous expansion of the scale of the power grid, the safety and reliability of the power grid operation has aroused widespread concern. Transmission lines are the main power components connected to the national power network. Since the transmission line crosses different complex terrains, the conductors are exposed to various environments for a long time, and it is very prone to failures such as strand opening. Once the transmission line fails, it will seriously affect the safe and reliable operation of the grid system. Therefore, the fault detection of transmission lines is one of the necessary procedures for the operation and maintenance of the national power network. [0003] The traditional power grid equipment moni...

AI Technical Summary

Problems solved by technology

[0004] In recent years, with the development of computer hardware and the rapid development of deep learning, the accuracy of image recognition and calculation time have improved, but the existing methods mostly use Faster R-CNN algorithm and complex networks such as VGG16 to combine equipment monitoring video and wireless Human-machine images are transmitted to the server for model training and fault identification of equipment faults. Large-scale networks and complex algorithms require the support of expensive hardware, which is not conducive to the promotion of equipment fault identification methods

Uploading the fault identification task to the server for identification, on the one hand, increases the load of data transmission, and also increases the computing power of the server. Traditional servers cannot meet the computing power of image recognition.

Traditional methods cannot meet the real-time requirements of transmission line fault detection

Images