| Road is an important part of human traffic, and it plays a significant role in the forest management. Forest roads are the necessary data source of vehicle navigation, emergency rescue, forest fire research, cadastral management, and GIS database updates, etc. With the rapid development of national economy, the demand on complete information of forest road is becoming larger. How to achieve forest roads information efficiently, timely, and accurately become the focus in information acquisition. The emergence of airborne LiDAR created a novel way to achieve forest roads. At present, the study of airborne LiDAR is mainly focused on improving the accuracy of point cloud, optimizing measure system, data filter, modeling buildings, and so on. In the aspect of filter, the outcomes of the studies are mainly suitable for flat areas, and not for mountainous regions; Using remote-sensing image to extract roads is an effective method, and the research on using point cloud datas from airborne LiDAR is not comprehensive. Considering the factors above-mentioned, achieving forest roads rapidly, accurately, efficiently is studied in this paper using the three-dimensional information and echo information. The concrete research content include:(1) The development of airborne LiDAR technology is summarized, and the components of airborne LiDAR system, the theory of measurement, data structures and main application of airborne LiDAR are introduced.(2) Several classical point cloud filters are analyzed, and brief evaluations of each filter. In this paper, the forest roads are extracted using the slope improved morphological filtering algorithm. That method is based on developed multi-scale morphological filter, suitable for steep area, and ensure the precision of forest road extraction.(3) This paper developed a classification method for forest road extraction from LiDAR point clouds through morphological features of forest road and Support Vector Machine (SVM). First, we deleted outliers and generated the digital surface model (DSM) and digital intensity model (DIM). Next, we achieved the DEM by means of the slope improved morphological filtering algorithm, and a potential road area can be obtained. Then, we classified the slope and intensity information of the potential road area by means of Support Vector Machine, and achieved the initial road areas that containing some small noises. The final road points can be extracted by filtering and refining the initial road areas with Morphological parameters.(4) Forest roads are extracted using the method proposed from this paper in the steep forest region by MATLAB programming. Results show that the prediction accuracy of this method is good, and the road extracted is nearly complete. In the end of the paper, some insufficient of the method are explained, and the research direction of forest road extraction in the future are mentioned. |
