Research And Implementation Of Aviation Surveillance Information Fusion Technology Based On Recurrent Neural Network

With the rapid growth of the number of civil aircraft and military aircraft,air traffic is becoming more and more busy.In order to ensure the flight safety of aircraft,the application of aviation surveillance information processing system is indispensable.The aviation surveillance information fusion technology in the system is to obtain accurate flight targets.The key to location information.Aviation surveillance information fusion is dedicated to merging multi-source detection data from the same target aircraft to obtain more accurate aircraft position,heading,acceleration,and more.Because the traditional aviation surveillance information fusion technology based on Kalman filter still has some shortcomings,such as poor integration in the maneuvering state,the model needs a lot of manpower and material resources to make repeated adjustments,etc.,so this paper uses Recurrent Netural Network(RNN).)Conducting aviation surveillance information fusion.This paper constructs a prototype system for aeronautical surveillance information fusion.The system consists of four parts:monitoring information preprocessing module,aircraft maneuver status discriminating module,aircraft position prediction module and multi-radar monitoring information fusion module.The monitoring information preprocessing module is responsible for monitoring message parsing,coordinate conversion and track matching.The aircraft maneuvering state discriminating module selects RNN as the model for training,and the trained model can judge whether the current aircraft is in a maneuver state based on a piece of track information.The aircraft position prediction module uses the weighted least squares method to fit the relationship between the monitoring information of each dimension and time.The fitted function can be used to predict the monitoring information of each dimension.The main purpose of aircraft position prediction is to The multi--source monitoring information of the target is aligned from different moments to the same time for fusion.The multi-radar monitoring information fusion module uses RNN as the model for training.The training process uses multi-radar data as input and ADS-B data as target output.In this paper,the track data is observed by the specific two radars to train the RNN model.In the maneuvering state discriminating module,when the RNN algorithm selects appropriate features,parameters and models,the accuracy of the maneuvering state judgment can be as high as 95.9%.In the aircraft position prediction module,when the weighted least squares method selects the appropriate weight function and parameters,the error of the single direction position prediction can be as small as 104 meters.In the multi-radar monitoring information fusion module,the track under multi-radar monitoring is used as test input data,and the ADS-B data is used as the target output data.The RNN-based multi-radar monitoring information fusion model can control the error to 202 meters,and this paper The average error of the radar data involved is 271 meters,so the use of RNN-based multi-radar monitoring information fusion technology reduces the error by nearly 70 meters.