Classification and Identification of Ship Types Based on Deep Forest
Aiming at the problems of single trajectory features,limited ship types of classification and lack of re-search on classification of marine ship types based on AIS data,a classification and recognition method of ship types is proposed based on Deep Forest.Firstly,16 motion features of the trajectory subsections are extracted to improve the diversity and identification of ship features.Secondly,the random forest is constructed as an individual learner to enhance the robustness of types classification using the resampling and selecting features randomly tech-niques of random forest.Finally,the four forests are integrated and connected in series to form a deep forest with the ability of adaptive iterative layer number,so as to enhance representational learning and generalization ability from depth.The experimental results show that the method is suitable for the classification and identification of ship types in small-scale data sets,and has a significant effect on the classification of yachts and other types of ships,which can effectively improve the accuracy of cargo and oil tankers and also have a high efficiency of classification.Moreover,there is room for improvement with the expansion of sample base and the optimization of model.
AIS datadeep forestmovement characteristicsship typesclassification and identification
万方数据
维普
