Research On Shipping Area Risk Model Based On AIS Big Data

The maritime transportation industry has stood out among the five major modes of transportation by its strong transportation capacity,good navigational conditions and excellent general performance since its inception.The researches for the navigational risk of vessels operating in sea and waterway areas,especially the decision-making analysis and methods of vessel collision risk,are greatly beneficial for maritime safety engineering and pollution preparedness and response planning.As the rapid development of the Automatic Identification System(AIS),applying the navigational big data to risk assessment and prediction has become an increasing trend.This thesis focuses on the models of regional vessel near miss collision risk based on AIS dig data.Our main contributions are as follows:Most recent studies have analyzed the risk of collision for a pair of vessels and propose micro-level risk models.The absence of a regional,open water vessel collision risk assessment system endangers maritime traffic and hampers safety management.This study proposes a new method for assessing regional vessel collision risk.This regional model considers spatial and temporal features of vessel trajectory in an open water area,the density complexity,a multi-vessel collision risk operator and possible relative striking positions.Finally,the k-means clustering method of multi-vessel encountering risk,based on the proposed model,is used to identify high-risk collision areas,which allows reliable and accurate analysis to aid implementation of safety measures.As experts are necessary to interpret the risk model results,this is time-consuming and resource-intensive.In order to mitigate this problem,a new method for recognizing regional ship encounter situations in terms of associated collision risk levels is proposed.In an attempt to obtain the risk levels of these encounters,this method applies Convolutional Neural Networks(CNNs),which is one of the most widely used algorithms in the field of deep learning,to process images of two-vessel or multi-vessel encounters.In addition,to enhance the expressive capability of the learned features,the AIS-derived near miss information is also introduced to serve auxiliary information.Our proposed method is able to fully exploit the powerful learning capacity of the deep neural network to effectively mimick the expert judgements of collision risk levels.The simulation results show that the new method can accurately classify vessel encounters according to their risk level,indicating that the modelling approach is a promising new direction to develop methods in support of maritime safety related decision making.Poor recognition and prediction of near miss collision risk can lead to catastrophic safety incidents for maritime safety engineering and pollution preparedness.Considering this problem,a nonlinear kalman Filter method for predicting ship encounter situations in terms of near miss collision risk is proposed.In this paper,the Extended Kalman Filter method is introduced,which provides the next optimal navigation data for ship-ship near miss collision risk analysis.Thereafter,according to the collision risk level,the near miss for the next two moments can be predicted by means of Unscented Kalman Filter method.The results indicate that the present work can accurately classify the collision risk level for two-vessel encounters and mitigate the human judgment error.