| Although GPS can provide high-precision three-dimensional coordinates, but as a result of GPS elevation, its relative to the earth reference ellipsoid high, rather than high-normal is usually required in relation to the quasi-geoid. Therefore, how will the GPS elevation converted to high-normal to become a focus of attention.Using the geometric method in the height anomaly fitting regional study, useing GPS and the height of level information, the most commonly used is a function model and function model based on stochastic models and integrated to be legal. Both are good fitting effect, but the accuracy of modeling and function model and the selection of stochastic models. Support Vector Machine (SVM) and Least Squares Support Vector Machine (LSSVM) theory for the data fitting provide a more novel approach. Based on the above research background, this paper has done the following:1. The function model of the multiquadric's central node to choose to do an in-depth discussion, put forward the "Adaptive location" to match the characteristics of the K-means clustering method. Experimental calculated results show that the abnormal elevation changes, whether flat or undulating region marked region, the fitting results are good.2. Based on the SVM theory, abnormal changes in the elevation of complex conditions, to improve the accuracy of SVM regression model conducted in-depth study of algorithms proposed amendment residual algorithm for support vector machine regression by fitting the experimental of mountain height anomaly, demonstrate the effectiveness of the algorithm.3. Choice of model parameters for LSSVM difficult problem, genetic algorithm (GA) into LSSVM Optimal choice of model parameters, the spreadsheet to avoid the cumbersome process of working to improve the computational efficiency.4. LSSVM combination of small sample, nonlinear data processing advantages, the "online least squares support vector machine regression" algorithm, in order to enhance the efficiency of the model solver.
|
PDF Full Text Request