A Research On Multi-Sensor Bias Estimation Under The Misassociations In Maritine Environment

With the rapid development of the national economy,exploitation and utilization of Marine resources are increasing.The large number of ships,which affects the traffic safety in the waters seriously,caused the frequent occurrence of marine accidents.The problem has aroused great concern all over the world.In order to ensure the safety of sea navigation,target tracking between ships and ships,ships and ports has become more and more important.Multi-sensor information fusion has many advantages in solving target detection and tracking problems and has become a key technology in the information age.The existing fusion system focuses on handling random noise and has difficulties in dealing with system deviations and track-association problems.Because of the complex marine environment,the system deviation of the sensor is likely to induce the false association of tracking,while the false association will affect the estimation accuracy of the system deviation.Although many scholars have done a lot of research on improving the tracking association,the false association of tracking can't be avoided under the influence of sea clutter and other factors.This dissertation focuses on the problem that the multi-sensor bias estimator can't work stably in the condition of false association as well as marine environment.The main work includes the following three parts:1)Discussion of basic issues in multi-sensor multi-target tracking.Based on the model establishment,simulation experiments are performed to reveal the necessity of system bias estimation in multi-sensor information fusion.The mechanism of the influence of miscorrelation on system estimation is analyzed theoretically,and the influence of miscorrelation on the bias estimation result is verified through simulation experiments.2)Robust Estimation of system bias under error associated conditions,the minimum median square estimator and Huber estimator are introduced,and an estimator that uses quantum particle swarm optimization algorithm and NM simplex algorithm for minimizing the objective function is proposed.Through the analysis of the objective function,it is verified that the anti-wild character of the least-squares method is better than the least-squares method.3)Aiming at the problem that the robust estimator mentioned above lacks precision in complex environments,a more robust system bias estimator is proposed by analyzing the algorithm.Firstly,in order to enhance the anti-wild characteristics of the estimator,WLAD-LMedS algorithm is proposed,combining the least-squares method with the minimum median square estimator;Then,ordering to solve the problem of the singular value appearing in the gradient of the objective function with a minimum multiplication when the residual is close to zero,Huber-LMedS algorithm is proposed.By analyzing the distribution of data,the feasibility of the proposed algorithm is verified.Based on work above,the algorithm is optimized to further improve its accuracy.Simulation experiments shows that the algorithm proposed largely inhibits the influence of miscorrelation on the estimation of bias.