Study On New Technology Of Vessel Tracking And Track Prediction In Vessel Traffic Services System

Vessel highly-precision positioning and real-time track prediction are the key technologies to realize VTS intelligent early-warning, and have strong application value. In view of the current methods for optimal ship plot estimation based on kalman filtering algorithm exist some doubts in model coordinate system choice, system running environment, noise background and handling of the filtering divergence, as well as some presented algorithms for track prediction are with poor practicability and universality. Thus, this paper makes a deep research, and the specific work is as follows:1.Considering the computation of predicted longitude in mid-latitude and Mercator method having derivation difficulties and the low precison problem, this paper puts forward a simplified, short-distance dead reckoning algorithm. Compared with the traditional algorithm of maritime navigation, the newly one has less parameters, and with high precision characteristics. Especially is suitable for some algorithms which need to build system state model, like kalman filtering.2.Considering the tranditional Sage-Husa adaptive filtering algorithm failing to estimate process and measurement noise simultaneously and the negative definite noise variance, this paper discusses the importance of measuring noise in low-dynamic system, and deduces a weight value range which guarantees the real-time estimated variance positive definite. Compared with original algorithm, this method owns less calculation step, better convergence, higher filtering precision and with non initial requirements.3.Considering the tranditional Sage-husa and STF adaptive kalman filtering algorithm cann’t estimate noise when in its divergence, this paper presents an improvement to solve it. By assignning a small part of the divergence to the measurement noise, balances the contribution to the kalman gain to conform to the actual engineering environment having unknown noise. This method owns less calculation step, good tracking performance, higher filtering precision, and with non initial requirements.4.In order to avoid the adaptive filtering algorithm have wrong measurement value associated problem under a single tracking source, this paper puts forward to a new method having a function of identification measuring value authenticity. By tracking the residual to identify the agreement of predictive value and measured value, then using state limit model to find which model turns to be abnormal. This algorithm could identify measurement value’s authenticity as well as having real time noise statistics estimation function. It owns less calculation step, better tracking performance, higher filtering precision, with non initial requirements and having the warning function of wrong associated measurement value.5.Considering some track prediction method have too many parameters and not able to fit the environment, this paper puts forward using BP neural network to predict ship’s track in the reality. To study the inherent rules from a group of continuous time interval ship dynamic data (longitude, latitude, heading, speed, time), sets up a mapping from the course and speed to the difference of longitude and the difference of latitude and then gets a ship motion law. The algorithm requires less for the parameters and is applicable to any ship. Especially suitable for the situation like hardly modeling ship motion process, and prefers general algorithm rather than for individual one. The experiments results show that it could capture the ship motion law within one second and has higher accuracy in prediction.