Lithology division for large-scale region segmentation based on LS-SVM and high resolution remote sensing images

Key words:  lithology division  large-scale region segmentation  least squares-support vector machines(LS-SVM)  high resolution  remote sensing

Abstract:

Based on the concept of large-scale region segmentation, extraction of texture, shape, spectral information of high resolution remote sensing image associated with the lithblogy and the advantages of least squares-support vector machines ( LS-SVM) in the non-linear prediction were used in the geological lithology identification. Firstly, the samples of spectral, texture, shape and altitude information which are relevant to litholo^ in the high resolution remote sensing images are selected. During the course of selecting, the image's texture is the main characteristic information. In the meanwhile, the chosen optimization feature space is based on the J-M distance and the degree of conversion classification. The feature space is compressed by using factor analysis and transformation dimension reduction, so that the characteristic information can be optimized. Hien, known samples are trained, and classification model is developed to evaluate model accuracy. Finally, the model was used to divide the study area's lithology and process classified objects. The classification method based on LS-SVM performs well in the high-resdution remote sensing images lithological identification，and provides a new method and means for the classification of geological lithology. LS-SVM classification model is more conducive in lithology identification after adding texture.