Improving the undulation estimation accuracy by Genetic Algorithm based Least Squares Support Vector Machine

摘要

Traditionally, the orthometric height H used in engineering application can be derived by leveling, which requires high cost of labor and time. On the other hand, the ellipsoidal height h derived by Global Positioning System (GPS) has the advantage of lower cost. The method of GPS Levelling can be applied to obtain the orthometric height from GPS-derived data. And for the transformation between ellipsoidal heights h and orthometric heights H, undulation N with sufficient accuracy is the main study goal. There exist a number of methods for approximating the undulation model. The polynomial method is the most widely used method to fit the geoidal undulation. However, the polynomial fitting method has its limitation when determined the undulation model in large areas with complex terrain. In order to improve the undulation estimation accuracy, the Genetic Algorithm (GA) is first used to search and optimize the parameters of LSSVM (i.e., LSSVM(GA)), and then use LSSVM(GA) to establish the undulation model. In this paper, 283 benchmark points distributed throughout the central part of Taiwan region with its orthometric height, ellipsoidal height and plane coordinates were used as test data. According to the test results, the accuracies of undulation estimation are improved about 42.83% (reduced from 0.0523m to 0.0299m) after using genetic algorithm based least squares support machine. The proposed method, LSSVM(GA), and test results will be presented in this paper.