SVM Based Constructing Method for Geomagnetic Reference Map

Geomagnetic reference map is the basis for geomagnetic navigation, and modeling of regional geomagnetic field is one of the available methods to construct the geomagnetic reference map. Common-used modeling methods are not so fit for the small-scale regional geomagnetic field, so new ways to modeling of regional geomagnetic field is needed. The total intensity of geomagnetic field is a complicated non-linear function of the longitude, the latitude and other factors. Since Support Vector Machine (SVM) is characterized by many excellent performances to approximate a complicated non-linear system, it is used to model the regional geomagnetic field in this paper. In SVM model, the longitude and the latitude are regarded as input variables, and the total intensity of geomagnetic field is regarded as the output variable. Of all the surveyed data, Some are used to train the SVM model, and the remain are used to verifying the precision of the model, and then the data needed to construct the geomagnetic reference map are predicted by the well-trained SVM model. Finally, the geomagnetic reference maps are constructed by both modeling method based on SVM and interpolation methods based on kriging and radial basis function (RBF), conclusion and discussion are given according to the experiment results.