Insulator defect detection method based on improved YOLOv5 convolutional neural network

Abstract

The invention discloses an insulator defect detection method based on an improved YOLOv5 convolutional neural network, which comprises the following steps of: firstly, improving a network structure of YOLOv5 from two aspects of reducing false detection similar objects and accelerating reasoning speed, and proposing the improved YOLOv5 convolutional neural network based on an attention mechanism and depth separable convolution; secondly, optimizing an algorithm loss function and post-processing aiming at the problem of insulator shielding caused by missing detection, calculating bounding box regression loss and a DIOU non-maximum suppression screening prediction frame by adopting a CIoU loss function, and further providing an ARS suppression algorithm based on an area ratio to reduce a multi-detection phenomenon of a defect target; and finally, training the improved YOLOv5 convolutional neural network to obtain a final detection network. According to the method, on the premise that the reasoning speed is not reduced, the target and the similar object can be correctly distinguished, and the shielded insulator can be prevented from being missed.

Description

technical field [0001] The invention belongs to the technical field of machine learning, and in particular relates to an insulator defect detection method. Background technique [0002] As an important part of the power network, transmission lines determine whether the entire power system can operate safely and stably. However, transmission lines have been exposed to the wild for a long time, and are affected by external factors such as high temperature, rain and snow, which are likely to cause failures such as aging, corrosion, and damage of transmission line components; Increase the possibility of aging and damage of line components, which will cause hidden dangers to the safety and stability of the power system. Once it occurs, it will cause heavy economic losses. Therefore, it is urgent to conduct fault inspections on transmission lines. Insulators are responsible for fixing current-carrying conductors in power transmission lines, preventing current from returning to gr...

AI Technical Summary

Problems solved by technology

Document 1 "Research on Defect Detection Method Based on Insulator Image [D]. Huazhong University of Science and Technology, 2019.", Document 2 "Insulator Fault Detection of Transmission Line Based on Deep Learning [D]. North China Electric Power University (Beijing), 2019." and Document 3 "Research on Power Insulator Fault Detection Method Based on Deep Learning [D]. Shaanxi University of Science and Technology, 2019." The Faster-RCNN algorithm was introduced into the insulator fault detection, and the fault location and identification were realized. The accuracy rate did not meet the ideal requirements. In addition, its reasoning speed could not achieve real-time performance; Document 4 "Research on Insulator Target Detection in Aerial Images [J]. Electric Measurement and Instrumentation, 2019, 56(05): 119-123." , Document 5 "Real-time detection of key components of power lines based on YOLOv3 [J]. Electronic Measurement Technology, 2019, 42(23): 173-178." and Document 6 "Recognition method of power components in UAV inspection images of transmission lines [D]. Chongqing University of Technology, 2020." Using YOLOv3 to achieve high-precision real-time detection, and through experimental data to prove that its average precision is not much different from Faster R-CNN, and the detection speed is not much improved

Images