Design And Implementation Of Collision Warning System Based On Ship Trajectory Prediction

In the context of economic globalization,the trade volume between countries is increasing day by day,and shipping,which has lower economic cost and can carry large-scale cargo,has become the main mode of transportation for export trade.However,this has led to the increasing density of ships in the sea,and the probability of collision between ships has increased greatly.Meanwhile,deep learning technology is increasingly applied to the field of ship collision warning,which has proved itself reliable in safeguarding the safety of ships and personnel and avoiding economic losses.Therefore,it is of great significance to study how to use deep learning technology to assess and warn the ship collision risk,replacing the manual warning which only relies on navigation experience,to ensure the safety of navigation.This paper addresses two main issues:first,due to the lack of foresight of most existing warning systems,the time left for crew members to react is very short,and the data transmission and computation on the sea surface led to a certain lag in the warning results.Secondly,current collision risk assessment algorithms usually do not include environmental factors into the assessment,and do not have a comprehensive assessment perspective on important factors such as nearest encounter distance and minimum encounter time.In this paper,based on the navigation data recorded by the automatic ship identification system,the residual CNN-LSTM ship trajectory prediction model based on the attention mechanism and the collision risk assessment model based on the fuzzy theory are proposed respectively,and the design and implementation of the collision warning system based on the ship trajectory prediction are completed on this basis.The trajectory prediction model introduces the residual block and attention mechanism to the base CNN-LSTM combined model,and the efficient Ranger optimizer is introduced into the training process of the model.Experimental results show that the prediction model proposed in this paper can achieve higher prediction accuracy compared with the basic CNN-LSTM model.The collision risk assessment model is based on the results of the ship trajectory prediction model,and the collision risk assessment is performed by the navigation data of ships at future moments,which well alleviates the problem of insufficient warning precognition and the lag of warning information caused by data transmission and computation.The model takes environmental factors into consideration compared with existing methods and establishes a more comprehensive and reasonable assessment model from both subjective and objective perspectives.The model constructs a two-indicator evaluation set and a five-indicator evaluation set for dynamic factors such as the distance of the nearest encounter point,and an environmental indicator evaluation set for static factors such as wind and wave level,and finally combines the weights of each indicator to obtain the collision risk assessment results.The experimental results show that the model has good performance in three aspects,namely,encounter potential,symmetry and model comparison.In the organization of the whole text,this paper firstly explains the research background and significance from the realistic value of ship collision warning,and analyzes the current situation in the field of ship collision warning through the relevant research results of domestic and foreign scholars.Then,this paper analyzes the requirements of the whole collision warning system based on ship trajectory prediction.Then,this paper presents the theoretical analysis,code implementation and model evaluation of the key problems that need to be solved,which are ship trajectory prediction and collision risk assessment.Secondly,this paper presents a detailed design of the overall architecture and functional modules of the ship collision warning system,and a detailed introduction of the classes and interfaces inside the modules and the interaction relationships under typical scenarios.Finally,this paper provides a specific description of the deployment details and test cases related to the system,and summarizes and outlooks the work of the ship trajectory prediction system designed in this study.