Based On CNN-BiLSTM Research And Application Of Multi-source Track Correlation Technology

In the field of modern command and control,information is an important factor that dominates situation assessment,and accurate acquisition of target global summary information is of great significance to command and control.In the field of information fusion,multi-source information fusion is an important aspect.The track correlation task can effectively discover the correlation between information and improve the effect of command and control.In the track association task,in view of the large deviation of track data collected by multiple sensors,this thesis realizes track data set processing through data preprocessing,feature engineering,data completion and similarity measurement.Aiming at the problem that the track duplication cannot be effectively removed due to the weak correlation effect,this thesis proposes the CNN-BiLSTM algorithm,which realizes the effective correlation of multi-source tracks in multi-scenario tasks by obtaining the spatio-temporal information of track data.The situational system based on the Spark framework is designed and completed,which improves the analysis and integration capabilities of the track data information of the track association module of the situational system.The main work includes:1.The track data set obtained by multiple sensors is processed.Utilization number processing,flight number upgrade for special expedition process operation,selection demand flight number adjustment,flight number adjustment for utilization internal and external projection method,similarity measure for base L1 flight number adjustment,final The number of flights in the model that can be Obtained by importing.2.A track association algorithm based on CNN-BiLSTM is presented.The track association algorithm based on CNN-BiLSTM can be used in the track association task of the situational system.The CNN convolution layer is used to model the space sequence of the track data,and the Bi-LSTM algorithm is used to model the time series to improve the track.The data features are expressive,and the resulting CNN-BiLSTM algorithm has a good use effect in multi-scene track association tasks and small track data association tasks.3.The improved CNN-BiLSTM model is used to the track association task,and the track association results are visualized.This article participates in and designs some functions of a situational system based on a distributed platform,realizes the inductive analysis of track-related information,introduces the design requirements of the situational system in detail,and demonstrates the overall design architecture and interface design of the situational system in detail.