Method and device for executing inspection task of unmanned aerial vehicle and electronic equipment

By using cross-modal semantic alignment and feature fusion to generate fused representation vectors, determine component instance information, generate flight control commands, and construct a topology graph, a technical closed loop for autonomous flight and topology information updating of UAV inspection is realized, improving the accuracy and efficiency of UAV inspection.

CN122450167APending Publication Date: 2026-07-24STATE GRID BEIJING ELECTRIC POWER CO
0 Cites 0 Cited by

Patent Information

Application Number
CN202610709390.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-05-21
Publication Date
2026-07-24

AI Technical Summary

Technical Problem

In UAV inspections, the lack of a unified embedding representation and cross-modal semantic alignment mechanism for multimodal sensor data leads to information silos, poor perception reliability, and a disconnect between perception recognition, flight control, and topology mapping, resulting in insufficient accuracy of autonomous tracking flight and lagging topology information updates.

Method used

By receiving multimodal sensor data, performing cross-modal semantic alignment and feature fusion, generating fused representation vectors, determining component instance information, generating flight control commands, and constructing a route space topology map, a technical closed loop of environmental perception, flight control, and topology mapping is achieved.

Benefits of technology

It improves the accuracy of UAVs' autonomous tracking flight over overhead lines and the efficiency of updating line topology information, solving the problems of low tracking flight accuracy and delayed topology information updates for UAVs.

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

Abstract

The invention discloses an execution method and device of an unmanned aerial vehicle inspection task and electronic equipment. The method comprises the following steps: receiving and responding to a predetermined inspection task, and obtaining multi-modal sensor data; executing cross-modal semantic alignment and feature fusion operation according to the multi-modal sensor data to obtain a fusion representation vector; determining component instance information according to the fusion representation vector; generating a first flight control instruction of the unmanned aerial vehicle according to the component instance information, and controlling the unmanned aerial vehicle to fly according to the first flight control instruction; and according to the visual imaging data, the space depth point cloud data, the attitude inertia sequence data, the position coordinate data and the component instance information, constructing a line space topological graph. According to the invention, the technical problems of low tracking flight precision of the unmanned aerial vehicle, lagging line topology information updating and lack of connection relation in the prior art are solved.
Need to check novelty before this filing date? Find Prior Art