A method for flight target trajectory classification based on bottleneck neural network embedding

By using a bottleneck neural network embedding method, features are extracted from flight target trajectory data and classified by a multi-voter system. This solves the problems of insufficient accuracy and robustness in flight target trajectory classification in traditional methods, and achieves more efficient recognition results.

CN116091813BActive Publication Date: 2025-12-1610TH RES INST OF CETC
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202211220036.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-10-08
Publication Date
2025-12-16
Estimated Expiration
2042-10-08

AI Technical Summary

Technical Problem

Traditional trajectory features are not ideal for representing flight targets, and popular target motion parameter features are easily affected by outliers, resulting in insufficient accuracy and robustness of flight target trajectory classification.

Method used

A bottleneck neural network-based embedding method is adopted to extract local and global features from flight target trajectory data, map them into low-dimensional embedding vectors through a bottleneck neural network, classify them using a multi-voteer, and finally identify them by combining support vector machine, random forest and XG-boost.

Benefits of technology

It improves the accuracy and robustness of flight target trajectory classification, simplifies the recognition model, reduces sensitivity to outliers, and improves fault tolerance and recognition speed.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116091813B_ABST
    Figure CN116091813B_ABST
Patent Text Reader

Abstract

The application discloses a flight target trajectory classification method based on bottleneck neural network embedding, and relates to the technical field of target identification, and comprises the following steps: firstly, extracting flight target motion features from flight target trajectory data; secondly, inputting the flight target motion features into a bottleneck neural network for feature embedding, and outputting embedding results; finally, inputting the embedding results into a multi-person voter composed of multiple classifiers, performing flight target trajectory type identification, and obtaining final identification results; according to the flight target trajectory data, the application mines target motion characteristics, and classifies flight targets by using the characteristics, so that the accuracy and robustness of flight target trajectory classification are improved.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Aircraft type recognition method based on target flight path

    CN108921219A

  • Residue soil truck monitoring method and system based on unmanned aerial vehicle and cloud platform

    CN114898236A