In this paper, remote sensing image classification method has been studied based on road trajectory prediction model, research background of remote sensing image road extraction has been reviewed, and key factors affecting road extraction have been analyzed, including image resolution and algorithm automation degree . Remote sensing image extraction methods of road elements, including the extraction strategies of fast road, main road and secondary road. In the aspect of fast path extraction, object-oriented classification and rule set construction technology are adopted. In main road extraction, the method combining the construction of cyclic rule set and mathematical morphology processing has been used. For sub-trunk extraction, the steps of image preprocessing and sub-trunk feature extraction have been proposed. These methods will provide effective technical support for road classification and extraction in remote sensing images.