| The information perception technology of the ship environment is the key part of intelligent ships,and the target location information on board is the basic data of ship information perception,which is also the key to realize the ubiquitous connection of ships,people and things.In recent years,passive indoor localization technology based on WiFi channel state information has attracted widespread attention.It can support intelligent applications such as real-time position management of ship personnel,monitoring of sensitive areas,and auxiliary emergency evacuation,etc.However,the current WiFi localization system for ordinary indoor environments cannot be directly applied to the ship environment.The dynamic characteristics of the ship and its special steel material will have a huge impact on the localization system.The purpose of this paper is to achieve high-precision and low-cost indoor personnel localization in the special environment of ships.To achieve this goal,we mainly carried out the following research work:(1)This paper studies the influence of ship environment on CSI and its channel parameters.First,extensive experiments were carried out on the real Yangtze River cruise ship,collecting a large amount of CSI data and its corresponding ground truth of volunteer’s location.After preprocessing the CSI data,the power delay profile algorithm,multi-signal classification algorithm,variance analysis and other methods are used for experimental analysis,and two major influencing factors are determined by the combination of experiment and theoretical analysis.We find that the dynamics of the ship’s environment lead to changes in the number of dominant multipaths of the signal,which further lead to channel parameter estimation errors.In addition,the steel material of ship can lead to loud noise and problems of confusion between metal reflectors such as bulkheads with personnel locations.(2)In order to solve the problem of the dynamic change of multipaths number in the dynamic ship environment,which leads to the decrease of the estimation accuracy of the channel parameters,a space alternating generalized expectation-maximization algorithm improved by the sparse variational Bayesian method is adopted.It can realize the joint estimation of the channel parameters and the number of multipaths.In the traditional space-alternating generalized expectation-maximization algorithm,the number of signal paths is set as a fixed value when estimating channel parameters,which is not in line with the characteristics of the ship’s environmental.The algorithm proposed in this paper incorporates the number of multipaths into the unknown parameter vector,and jointly iteratively estimates the number of multipaths and channel parameters to realize the estimation of channel parameters adaptive to the ship environment.We evaluate the estimation accuracy and cost of the algorithm,and the results show that the algorithm can achieve good results.(3)Aiming at the problems of loud noise and the confusion of the locations of metal reflectors and personnel caused by the steel material of ships,a convolutional recurrent neural network localization model is proposed in this paper.First,the estimated multi-dimensional channel parameters are converted into a two-dimensional likelihood matrix containing key information,which is used as the input data of the neural network.Using the characteristics of the continuity of person ’s location changes and the different movement pattern of reflectors such as person and walls,the neural network automatically identifies the parameters related to person and then mapping them to the person’s location,and finally obtains the continuous position of the person.The experimental results show that the median localization accuracy of the system proposed in this paper reaches 0.92 m,which can meet the needs of personnel localization in the ship environment. |
