Datasets
Standard Dataset
Track data
- Citation Author(s):
- Submitted by:
- zekai wu
- Last updated:
- Wed, 02/26/2025 - 09:06
- DOI:
- 10.21227/5wm7-gw84
- License:
80 Views- Categories:
- Keywords:
Abstract
The ADS-B track data, downloaded from the open Internet, contains four aircraft targets, namely B737, C17-1, C17-2, E35L, each target contains 110 complete tracks, totaling more than 100,000 track points. The data sets of 4 aircraft targets were randomly divided into training sets, verification sets and test sets according to the ratio of 8:1:1. The track used for training in the experiment is stored in the form of track points, using UTF-8 encoding, and each line corresponds to a track point at one time, the contents are respectively time, longitude, latitude, altitude, speed, heading and climb. All data is stored in txt format.
Aircraft individual recognition can greatly enhance the perception of battlefield situation, but it also faces many problems, such as multi-source features are limited by conditions, only single source features can be obtained, and there are few deep analysis based on measured track data as individual identification features.Based on this, this paper proposes an aircraft individual GRU recognition algorithm based on multidimensional track feature data.Firstly, the feasibility of using six-dimensional track data of longitude, latitude, altitude, speed, climb rate and heading angle as aircraft individual identification features is analyzed from the perspective of dynamics and geographic information space. Then, based on the Area Under the Curve (AUC), a set of evaluation model for track feature selection is established. Finally, a Gate Recurrent Unit (GRU) aircraft individual recognition algorithm combined with Attention Mechanism is constructed based on the selected track features.Based on the original measured track data set containing 4 aircraft individuals, this paper proves the correctness and effectiveness of the selected track features used in the evaluation model of aircraft individual identification and track feature selection by visualization analysis and AUC calculation, and uses the constructed GRU-Attention model for individual identification of the measured unknown track. The results show that the recognition accuracy is 71.64%, which is higher than other traditional classifiers. The effectiveness of the algorithm has been verified, and the aircraft individual recognition has been realized to a certain extent.
