A tree obstacle detection method, detection device and computer equipment

By expanding the tree barrier image training samples and performing fusion processing of the style migration network, the problem of low tree barrier detection accuracy in the prior art is solved, and higher detection accuracy is achieved.

CN114219950BActive Publication Date: 2025-03-28JIANGSU FRONTIER ELECTRIC TECH
2 Cites 0 Cited by

Patent Information

Application Number
CN202111485336.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-12-07
Publication Date
2025-03-28
Estimated Expiration
2041-12-07

AI Technical Summary

Technical Problem

The existing tree barrier detection methods have low detection accuracy due to the limited number of training samples and the difference in sample environment.

Method used

By expanding the number of tree barrier image training samples, the initial tree barrier detection neural network is used to detect the historical images of the drone patrol power lines, obtain the second tree barrier image training samples, and extract content information and style information through the style migration network to form the third tree barrier training sample, and perform secondary training to improve detection accuracy.

Benefits of technology

By expanding the number of training samples and improving the fit of the sample environment, the detection accuracy of the tree barrier detection neural network is significantly improved.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present application provides a tree barrier detection method, detection device and computer equipment, the method comprising: inputting a first tree barrier image training sample into an initial tree barrier detection neural network for training, obtaining a first tree barrier detection neural network for detecting tree barriers; obtaining historical images of power lines inspected by drones, detecting the historical images through the first tree barrier detection neural network, obtaining a second tree barrier image training sample; using a target style transfer network to fuse the content information of the second tree barrier image training sample with the style information of the first tree barrier image training sample, obtaining a third tree barrier image training sample; inputting the second tree barrier image training sample and the third tree barrier image training sample into the first tree barrier detection neural network for training, obtaining a second tree barrier detection neural network, and using the second tree barrier detection neural network to detect the real-time image of the power lines inspected by drones, obtaining a target tree barrier image. The accuracy of tree barrier detection can be improved.
Need to check novelty before this filing date? Find Prior Art