Research And Application On Practical Constrained Principal Curve Algorithm

ABSTRACT:Accurate GPS digital map plays an important role in train control system for the train safety, positioning, scheduling, and in highway vehicle navigation, route selection, induction and other aspects. However, the accuracy of GPS receivers will be affected by a variety of errors, and the vehicle system itself has a very complex electromagnetic environment, and therefore, GPS receivers measure the data is difficult to achieve high precision. If using the differential GPS system, the cost will increase exponentially. Furthermore the data of GPS digital map is enormous. Storing and using these data not only need huge storage space but also increase the running time.Principal Curve algorithm proposed in 1984 by Hastie. After a larger development, it has a great performance in image recognition, trajectory control, data reduction and other issues. In this article, principal curve algorithm has been research and improvement. It can increase the accuracy of measurement, reducing storage space, running time and costs in the generation of GPS digital maps which has an important application value.In this article, K principal curves based on the theory of principal curve algorithm generate and optimization methods were further studied and improved firstly. And practical principal curve algorithms have been proposed including the three optimization method:ALL optimization (incessant increasing number of the vertex and making global optimization algorithm), MPM optimization (two steps and using max point optimization algorithm) and DC optimization (using the split zoning first and then combined by merging algorithm).Then software has been developed by using Matlab and C#. After that we did the research on the computing error, the fitness and robust of algorithm and other aspects by simulation data through simulation experiments and did the research on the storage space and computing efficiency and other practical aspects by railway and highway GPS data. The results show that the proposed principal curve algorithm has good performance in the calculation of accuracy, computing efficiency, data storage, fitness and robust.