Douglas Peucker algorithm extraction and deep learning classification of ship wandering trajectory

Ship loitering behavior is a kind of ship motion that frequently turns in local space. The ship loitering pattern is related to ship motion intention, and has important research significance for ship type recognition and abnormal behavior detection. Aiming at the problems of the extraction efficiency and accuracy of the existing ship loitering trajectory extraction algorithms, and the need to improve the classification accuracy of loitering trajectories, a ship loitering trajectory extraction method based on Douglas Peucker algorithm is designed, and the concept of loitering wave point is proposed to define the large steering change of the ship. At the transition point, the extracted wandering trajectory data is converted into a wandering trajectory image dataset, and the wandering trajectory is divided into four forms: disordered round-trip, two-point round-trip, forward round-trip, and cluttered line group. And a ResNet50 deep learning model is constructed to train ship wandering trajectory data set, and the Adam algorithm is used to optimize the model, which improves the training efficiency of the model. The experimental results show that the precision of the algorithm of ship wandering trajectory extraction can reach 98.56%, and the wandering trajectory extraction efficiency is improved by more than 23%. In the recognition and classification of the four wandering trajectory morphology data sets, the model obtains an average accuracy rate of 91.03%, which is better than the VGG16 model and SVM（Support Vector Machine） model in the contrast experiment.