Ship Track Prediction Based On CNN And LSTM

Water transport is the main means of transport for economic and trade exchanges among countries.It has the characteristics of large carrying capacity,low cost and wide range.In recent years,water traffic accidents,smuggling activities and other conditions show a rising trend,the security situation is not optimistic.It is necessary to predict the track of the target ship according to the known historical position information,so as to deduce the position information of the ship in the future for a period of time,in order to better supervise the navigation condition of the waters and avoid the accidents such as collision,grounding and running aground,therefore,it is very important to predict the ship’s course accurately and effectively.In the research of ship track prediction,deep learning prediction method is widely used,which is a kind of algorithm based on artificial neural network to represent the data,and has nonlinear,self-adaptive and self-learning ability.Due to the real-time requirement of ship track prediction,most of the prediction algorithms are based on a single model,such as convolved neural Networks(CNN),LSTM Networks-long Short Term Memory(LSTM),etc.,at present,due to the increasingly complex water traffic environment,the single model is difficult to meet the accuracy requirements,and how to solve the contradiction between the real-time and accuracy of ship track prediction is a challenge.CNN-LSTM is the most widely used hybrid model algorithm for the prediction of the correlation of things before and after,which combines the advantages of CNN and LSTM to make the identification or prediction more accurate,it is widely used in speech recognition,human behavior recognition and text image classification.In this paper,a hybrid model Algorithm based on ship track prediction is designed,in order to solve the contradiction between real-time performance and accuracy of single model in ship track prediction,the real-time performance of the model is improved by means of AIS input data block label,Adam function optimization and Dropout preventing over-fitting.Finally,the simulation results with the AIS trajectory data of the Yangtze River cargo ship show that the combined model based on One-dimensional CNN-LSTM is better than the single model,and the running time is shorter than that of the LSTM model,which is close to the CNN model,but the accuracy is due to the CNN model.