Research On Prediction Method Of Ship Track Based On Multi-source Heterogeneous Data

With the prosperity of maritime trade ship traffic density gradually increases,port waters become more crowded,and the problem of ship traffic safety becomes more and more prominent.In order to achieve better maritime navigation supervision,ship trajectory prediction has become a hot spot in the field of research.At present,most of the existing researches are based on AIS data for ship trajectory prediction,and there are also a few predictions combining AIS and radar image data,but all of them have a prediction time range of seconds,which is difficult to adapt to the increasingly complex maritime traffic situation.Therefore,a ship trajectory prediction method based on multi-source heterogeneous ship trajectory data is proposed to achieve relatively longer time trajectory prediction of ships in port waters.Firstly,a distributed ship trajectory database is constructed,and the integration of multi-source heterogeneous maritime data is realized based on data integration theory;on this basis,a ship trajectory state mining algorithm is designed,and the ship trajectory prediction model is constructed using the ship trajectory state mining results,and then the model parameters are solved based on the additive Viterbi algorithm;finally,the prediction algorithm is verified with the ship trajectory data of Qingdao port,and the experimental The experimental results show that the proposed ship trajectory prediction model can predict the ship trajectory in a relatively long time range,and the prediction results have good interpretability and the prediction method has strong applicability.The innovation points of this study are summarized as follows.(1)A dynamic route-guarded ship trajectory extraction framework is proposed for improving the efficiency of reading and writing ship trajectories under high concurrency scenarios.The ship trajectory segmentation algorithm is designed based on Redis cache,which supports pluggable trajectory segmentation rules.Combining with ZSET data structure to achieve efficient caching of ship trajectory segments,the ship trajectory memory-based calculation is realized,and the efficiency of trajectory calculation and analysis is improved.(2)A distributed ship navigation database based on Mongo DB is constructed to realize the integration of ship trajectory data storage and multi-source heterogeneous maritime data.The integration method of multi-source heterogeneous data(AIS,ship archives,vector electronic nautical charts and hydro-meteorological data)is designed according to the data integrity principle,and then a spatio-temporal index is designed and a distributed ship navigation database is built on the basis of Mongo DB to provide a data analysis basis for the subsequent prediction model construction.(3)A ship trajectory prediction model based on Hidden Markov Model(HMM)is designed to realize ship trajectory prediction in a relatively long time range.Firstly,an incremental state mining algorithm based on the ship trajectory cache is designed to construct the ship trajectory state set and then build the HMM-based ship trajectory prediction model,then the trajectory prediction is realized based on the additive Viterbi algorithm,and finally the model is verified by using the multi-source heterogeneous ship navigation data of Qingdao port.