Road Extraction Based on the Algorithms of MRF and Hybrid Model of SVM and FCM

Information extraction is the prerequisite of remote sensing image segmentation, which is the key procedure of image analysis. In this paper we establish MAP-MRF framework using Markov random field (MRF),adopt the sampler training, and get the factor of model, introducing the simulated annealing to segment the image and extract the road. On the other hand, Support Vector Machines (SVM) plus Fuzzy C-Mean (FCM) model was proposed and integrated together for remote sensing image segmentation. Firstly, we need use non-supervised clustering for remote sensing image by using FCM; then SVM is adopted for further classification, and extract the road. Finally the comparison with two proposed algorithm was carried out, and after experiment, SVM plus FCM model is much more accurate than Markov random fields.