Application Research Of Multi-source Information Fusion Technology In Water Transport

At present, in the field of water transportation, monitoring the state of navigation mainly based on radar, AIS system, GPS system, VHF communication system and CCTV system. in the VTS, system, the location information and speed information of the ship is usually based on radar or GPS. In practice, using a single or the same type of sensors to detect target has problem more or less, such as information unitary and low accuracy of information. Some of the existing studies have shown that different types of sensor using a reasonable fusion model, not only enhanced ability to detect targets on space, but also can be extended to detect target on time span.AIS has many advantages, for example, it is more comprehensive collection of information, and the location data and speed data of the mobile target detected by AIS is more accurate, and it not be easily affected by topography, weather and light. Therefore, AIS can provide a more accurate and reliable ship's navigation information. However, AIS also has some disadvantages. For example, ship's position and speed information detected by AIS is really collected by the GPS, but GPS is easily interfered by human; and GPS receiver is easily to lose lock on the navigation satellite in high noise conditions; and it also has multipath interference and ionospheric delay and other problems; more importantly, AIS can only detects the targets that have installed AIS equipment. At present, not all of the ships are equipped with AIS equipment, especially inland ships. Radar can detect river shoreline, buoy, and most of the objects. However, radar is also easily affected by weather condition, electromagnetic, topographic and other factors, and it easily detect water birds, water waves, rocks and other clutter points, and it particularly difficult to distinguish the two goals that are too close, and the information detected by radar is single. Analyzing the characteristics of AIS and radar, AIS and radar can make up some shortcomings of each other, but can not completely replace each other. Therefore, integration between AIS, radar, CCTV and other devices, and the fusion application of information collected by a variety of sensors is a very hot research for long time, and it will be the direction of development of intelligent shipping management. At present, the research of multi-source multi-sensor information fusion focused on the military field, such as research of AIS and radar information fusion, research of AIS and integrated bridge system information fusion, them opens up a precedent and gives a direction for depth and comprehensive multi-source information fusion study. In civilian areas, especially in the field of inland water transport management, it is necessary to make further research on multi-source information fusion. In this case, any research of information fusion will promote inland water transport management, and it will play an active role for the safe navigation of ship and the development of economy of inland water transport.In this thesis, AIS and radar target position and velocity data fusion and the ship's track prediction is preliminarily studied. Throughout the study, three key researches are done mainly. Firstly, research about time registration of the target ship's position and speed data detected by AIS and radar; secondly, research about the data fusion of the same target ship's location data and speed of data; thirdly, research about the ship's track forecasting. Some new methods of data fusion based on the existing fusion model about AIS and radar data is studied. For example, RBF neural network interpolation algorithm is used for time registration of various types of sensor data. And method of data fusion is improved in this thesis, such as the GA-PSO algorithm that is used to data weighted fusion. Finally, GA-BP is used to predict the ship track.Throughout the course of the study, all multi-source data-processing models and algorithms are tested through MATLAB simulation. The simulation results show that the new data-processing models and methods are more accurate, faster, better than the existing models and methods, and them can be directly embedded into the existing system and have some practical value.