Research On Intelligent Analysis Of Ships' Behavior Characteristics Of Entering And Leaving Port

It is dangerous when ships sail into and away from port waters because of the complicated navigation environment and the limited waters.It is also affected by wind,waves,currents,and pilots' unfamiliarity with the port's hydrological environment,which increases the risk of safety accidents in port waters.At the same time,the efficiency of ships entering and leaving port is low,which in turn affects the port's production efficiency.To grasp the navigation situation of ships in port waters in real-time,improve the level of intelligence of ports and ships,and guarantee the navigation safety of ships.This thesis takes ships sailing into and away from port as the research objects,and aims to help the regulatory authorities to supervise the ships in the waters of the jurisdiction more effectively.This thesis has completed the following main work:Data preprocessing is a very important part of the data mining process,which greatly affects the efficiency of data mining algorithms.Through the analysis of the original ship trajectory data set,it is found that there are problems such as field errors and obvious noise,and it is irregular time series data.Therefore,this thesis designs a simple and efficient data processing algorithm to improve the accuracy and completeness of the data,and effectively reduce the amount of calculation in the data processing process.Preprocessing the original data,removing the noise data,adjusting the storage format of the data to meet the input requirements of the model.Based on Hausdorff distance,an improved algorithm for trajectory similarity measurement with direction resolution is proposed.Because of the problem that the direction of the trajectory cannot be identified in the process of trajectory clustering,a measurement method of superimposing the cosine distance based on the traditional Hausdorff distance is proposed to enable it to smoothly distinguish ship trajectories of the same shape and reverse direction,thereby improving the accuracy of trajectory clustering.A trajectory simplification algorithm with course consistency and adaptive compression ratio is proposed.The algorithm simplifies the trajectory by using the adaptive compression ratio method through the preset compression effect of the user and finally outputs the simplified trajectory,which can effectively reduce the calculation amount of the trajectory clustering process.The experimental results show that the clustering algorithm proposed in this thesis can distinguish the ship's trajectory direction and reduce the computational cost of the clustering process by simplifying the trajectories.Predicting the changing trend of the navigation environment is an indispensable part of the field of environmental perception.Especially in crowded port waters,it is particularly important to predict the movement trend of the target ship.Aiming at the problem of low accuracy in the long-term prediction of ship trajectories by existing methods,this thesis proposes a ship trajectory prediction method based on long-short term memory neural network(LSTM),which improves the accuracy of long-term prediction of ship trajectories.This thesis collects a massive ship's historical trajectory data and obtains the ship's state information at every moment.Based on the traditional recurrent neural network,long and short term memory units are added to build a ship trajectory prediction model.This task is regarded as the sequence of the ship's position in space in a specific time dimension,and the ship's position information probability distribution at the future time is output by observing the ship at the first few moments.The model is trained and tested using the ground-truth history AIS trajectory data sets.The experimental data shows that the model in this thesis is better than the existing methods in predicting the ship's long-term position by learning the ship's historical motion data.To improve the safety of ship navigation and more effectively monitor ships sailing in port waters,it is necessary to track the ship and identify whether the ship's sailing behavior is abnormal in real-time.The thesis proposes a method for detecting abnormal ship behavior based on the probabilistic directed graph model.Based on the clustering of ship trajectories,for each type of trajectory,the instantaneous state of the ship is discretized into the corresponding grid.Taking the grid as the basic unit,the statistical characteristics of the historical trajectory of ships in the port in the geographical space are obtained.Using the grids as nodes,a trajectory directed graph is established.Through the study of the ship's historical trajectory data,the weights between the nodes in the graph are obtained.Calculate the mean and variance of the ship's heading and speed dimensions in each node to get its statistical characteristics,and build a ship state abnormality detection model based on the obtained statistical characteristics of the data.In the probabilistic directed graph model,given the current node,a ship position anomaly detection model is constructed by searching all the child nodes of the node and calculating the conditional probability of each child node.This thesis uses the navigation simulator platform to design ship navigation experiments to verify the performance of the model.Four sets of ship simulation and manipulation experiments were designed for different ship types.The experimental results show that the total time consumption of the model does not increase significantly with the increase in the number of detection tasks,indicating that the model is capable of real-time detection of ship behaviors during navigation.The model detects the ship's speed,course,and other abnormal status on average accuracy is 93.35%,and the average accuracy for detection of ship potential anomalies is 92.68%.It is proved that this model can detect most abnormal ship behavior,and it is superior to the existing methods in the commonness of abnormal behavior detection.