Research on application of neural network to GPS coordinates transformation
Han Min · Dalian Ligong Daxue xuebao · 2005
The integration technology of Global Positioning System (GPS) and Geographic Information System (GIS) is one of the key technologies emerging in the wetland study. The neural network method is applied to the research on coordinates transformation method between local map coordinates system and global coordinates system of GPS data. A three-layer feed-forward neural network model is built to transform data. Levenberg-Marquardt algorithm is used in the transformation process, and a modified performance function is adopted in neural network training to improve the network generalization. Neural network model, principle of learning algorithm, simulation curves and output results are discussed in detail. The simulation results prove that Levenberg-Marquardt algorithm is suitable to the GPS data coordinates transformation. The study is undertaken on Songliao valley and Zhalong wetland in Heilongjiang Province which is taken as the central study area. The result of practice proves that neural network method of transformation succeeds in building the complex model of transformation and is an effective method with high precision in dynamic data analysis of wetland protection.