Research On Multi-sensor Target Track Association Based On Machine Learning

With the rapid development of Internet and electronic technology,the situation estimation of maneuvering target has been widely used in many fields,and track association is one of the key problems to solve the situation estimation of maneuvering target.Track association is the simultaneous interpreting of two tracks from different sensors to represent the same target.With the increasing number of sensors for measuring targets,the complex situation of target track crossing,bifurcation,etc.makes the track correlation process difficult.In view of the disadvantages of the original track association algorithm,such as complicated calculation process and low correct association rate,this paper will use the algorithm in machine learning to solve the problem of track association.The specific work includes the following: 1.Firstly,several kinds of common maneuvering target motion states are introduced.The interactive multi model algorithm is mainly studied,which is used to generate the overall motion state trajectory of maneuvering target in a period of time.And the interactive multi model algorithm is used to simulate and analyze the maneuvering target.In addition,several traditional track association algorithms are studied: weighted track association algorithm,modified track association algorithm,nearest neighbor track association algorithm.The idea of several algorithms to solve the problem of track association is analyzed.The shortcomings of these algorithms in the case of multi-target multi-sensor and track crossing are described.2.Because the traditional track association algorithm is to judge the correlation between two tracks,when the number of targets is large and there are a large number of tracks,it is a problem that the processing ability of the traditional method is insufficient,which leads to the decline of correlation accuracy.By using the clustering analysis algorithm in machine learning to deal with the problem of track association,aiming at the problem of track association,this paper mainly studies the K-means clustering algorithm in machine learning.Aiming at the problem that the clustering result of track data in k-means algorithm is fuzzy and uncertain in the process of track association,the improved algorithm fuzzy k-means(fuzzy)is studied K-means)to cluster the track data.Considering that the current state of maneuvering target is related to the historical state,but the correlation of track data in time sequence is not considered in the fuzzy k-means algorithm,an improved algorithm SFKM is proposed based on the fuzzy k-means algorithm,which considers the correlation of track data in time and improves the accuracy of track correlation through simulation.3.Although SFKM algorithm improves the accuracy of track association,with the increasing number of clusters,the association accuracy decreases gradually.To solve this problem,this paper uses convolutional neural network(CNN)and long-term memory network(LSTM)in deep learning to extract features of track data from spatial dimension and time dimension respectively,so as to improve the feature expression ability of track data At the same time,it improves the intelligence and accuracy of track association in sensor data processing.The simulation results show that the proposed method achieves higher correlation accuracy than the original track association algorithm mentioned in this paper.