Research On Multi-sensor Data Fusion Algorithm For Ship Target Detection

Vessels monitoring plays a very important role in ocean management and maritime safety.Synthetic Aperture Radar(SAR),High Frequency Surface Wave Radar(HFSWR)and Automatic Identification System(AIS)are three common sensors used to monitor vessels.Each of the three sensors has its own characteristics.Multi-sensor data fusion can improve the detection accuracy of the vessel and expand the detection range.In addition,it can also provide more information to maintain the maritime rights and interests for the country.This article focuses on the fusion detection of ship target for HFSWR,AIS,and SAR,and conducts research on point-to-point association,track generation and track-to-track association.The research content is divided into the following points:1.A global optimal point association algorithm based on Munkres is proposed for SAR and AIS point-to-point association of vessel targets.The cost function is established according to the feature attributes,and then the cost matrix is calculated.The point association results are obtained according to the optimal allocation of the cost matrix,and the point association result is verified according to Euclidean distance.The experimental results show that when the association accuracy is basically the same,the point-to-point association result obtained by this algorithm is better than point-to-point association result of the nearest neighbor.2.An algorithm is designed to generate a track for vessel targets with unknown observation noise.Firstly,a double iterative variable Bayesian adaptive Kalman filter algorithm is used to predict the state of the vessel target with unknown observation noise.Secondly,a double-threshold nearest neighbor algorithm is used to perform point-to-track association.Finally,the track is corrected according to the statistical distance.The experimental results show that the designed algorithm can well generate trajectories for vessel targets monitored by HFSWR,and can remove the incorrectly associated point from the generated trajectories to get more correct trajectories.3.A global optimal assignment track association algorithm is designed for the HFSWR and AIS track association problem with cross distribution and dense distribution,and an association confidence calculation method is designed to verify the validity of the track association results.The measured HFSWR and AIS trajectory data indicate that the algorithm can effectively improve the accuracy of the trajectory association compared with the mean nearest neighbor algorithm when the trajectories are crossed and densely distributed,and the confidence of the track association is higher than that of the mean nearest neighbor trajectory association.