A Method and System for Anomaly Detection of UAV Atmospheric Data Based on Spectral Analysis

By using spectral analysis to decompose and detect multi-scale disturbances in UAV atmospheric data, this method solves the problem of distinguishing disturbance components and locating composite anomalies in existing technologies, achieving high-precision anomaly detection and correction, and improving the quality of UAV atmospheric observation data.

CN122087670BActive Publication Date: 2026-06-30NANJING TIANQING AEROSPACE TECH CO LTD
2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
NANJING TIANQING AEROSPACE TECH CO LTD
Filing Date
2026-04-23
Publication Date
2026-06-30

AI Technical Summary

Technical Problem

Existing methods for detecting anomalies in UAV atmospheric data are unable to effectively distinguish between disturbances from different sources, making it difficult to locate complex anomalies and resulting in a lack of targeted dynamic correction, which affects the accuracy of the correction.

Method used

A graph-based analysis approach is adopted. By constructing dynamic and motion condition vectors, a time-frequency distribution matrix driven by dual condition vectors is introduced for multi-scale perturbation decomposition. Anomaly detection and root cause localization are performed by combining graph convolution and multi-head attention mechanisms. Observation layer and anomaly layer graph models are constructed for dynamic correction.

Benefits of technology

It enables the effective identification of multi-scale disturbance components and accurate location of complex anomalies in UAV atmospheric data, improves the accuracy of anomaly detection and the pertinence of correction, and ensures the quality and reliability of atmospheric observation data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122087670B_ABST
    Figure CN122087670B_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for anomaly detection in UAV atmospheric data based on graph analysis, belonging to the field of data detection technology. The method involves real-time acquisition of atmospheric observation data, time synchronization processing, and spatial mapping processing to generate an atmospheric data sequence. Time-frequency analysis is performed on the atmospheric data sequence, introducing a time-frequency distribution matrix driven by dual-condition vectors to decompose the atmospheric data sequence into multi-scale perturbation vectors, generating multi-scale perturbation vectors. These multi-scale perturbation vectors are output to a preset anomaly graph detection model. Adaptive learning of the graph structure and a multi-head attention mechanism are used to propagate and fuse node features, identifying single and compound anomalies, inferring the root causes of the anomalies, and outputting the anomaly detection results and root causes. Based on the anomaly detection results and root causes, targeted corrections are made to different observed anomalies to obtain corrected atmospheric data, thereby improving the quality and reliability of UAV atmospheric observation data.
Need to check novelty before this filing date? Find Prior Art