The poor quality of real ship data and the lack of navigation status marks make it different from the experimental data recorded with experimental conditions, which is difficult to be effectively used in the study of ship performance. In this paper, a visual analysis system of navigation logic is developed to assist in abnormal data processing. At the same time, the visual analysis is used to divide the real ship data sub units, and the small sample navigation state determination set is obtained. The results of K-means clustering are compared with the determined set, and the navigation state class is obtained by semantic classification. Factor analysis is used to analyze the characteristics of navigation state data, so as to reduce the dimension of data and make the classification result better. Taking the inland navigation mark maintenance work boat data as an example, this method is realized to label the ship navigation status for the unknown large sample data set.
Ship's navigation state reflects the captain's thought of steering the ship. In this paper, the captain's thought of steering is defined as the captain's response to the operating requirements (such as constant speed cruise or docking).By executing a series of actions (such as decelerating or steering), the ship reaches a certain sailing state (such as straight sailing or berthing).Therefore, there is no doubt that the real ship data contains rich ship behavior characteristic information such as the captain's experience in operating the ship, the movement and maneuvering characteristics of the real ship, and etc. However, there are many difficulties in collecting relevant information from the real ship data through monitoring systems such as AIS. One problem is the unimprovable data quality of the actual ship's data which is used in data analyze. For data analyzing, data and methods are equally important and closely related. When data quality is poor, no matter how good the analysis method is, it might be unable to carry out accurate and reliable results.Another important point is that these data is not connected to the sailing environmental conditions and the captain's operation considerations. This is also one of the biggest differences between the actual ship data and the experimental data. The experimental data is connected with the sailing conditions and ship operation thoughts while the actual ship data is not, so it is impossible to tell what the ship was doing at that time, whether it was berthing,maneuvering, or sailing straight from the data.