Defective insulator identification method based on YOLOv3 network and particle filter algorithm

Abstract

The invention relates to a defective insulator identification method based on a YOLOv3 network and a particle filter algorithm. According to the defective insulator identification method based on a YOLOv3 network and a particle filter algorithm, the defect of an insulator in an inspection image containing a complex background in a real environment is identified based on a YOLOv3 network and a particle filter algorithm. The efficiency of insulator defect identification is greatly improved, and a reliable technical support is provided for the power grid maintenance department.

Description

technical field [0001] The invention relates to a defective insulator identification method based on a YOLOv3 network and a particle filter algorithm, and belongs to the technical field of line maintenance of smart grids. Background technique [0002] With the rapid development of the national economy, people's dependence on and demand for electric energy is also increasing, and the resulting power inspection tasks are also increasing. In recent years, the improvement of the level of science and technology has gradually replaced the traditional manual line inspection methods such as drone line inspection and robot line inspection. These emerging line inspection methods have improved work efficiency and safety performance to a certain extent. However, a large number of inspection images with complex backgrounds and a wide variety of defects still pose great challenges to the relevant maintenance departments. With the continuous development and maturity of artificial intellig...

AI Technical Summary

Problems solved by technology

[0004] At present, most insulator defect recognition algorithms based on inspection images use basic image processing or pattern recognition methods to detect insulator defects. This type of algorithm requires a relatively pure image background and high contrast between the insulator and the background; Introduced into the detection process of insulators, but most of them stay in the identification stage of insulators. There are two problems in directly using deep learning to detect insulator defects: one is that the defect samples are seriously insufficient, and it is difficult to support the training of the network; high robustness

However, the defects in insulators are often very small, and there are few samples of insulator defects at present, and YOLOv3 alone cannot complete high-accuracy detection

Images