Tree detection in LiDAR data

摘要: In this paper, we present an approach to detecting trees in registered aerial image and range data obtained via LiDAR. The motivation for problem comes from automated 3D city modeling, which such is used generate the models. Representing these models problematic because are usually too sparsely sampled tree regions create accurate 3-D model of trees. Furthermore, including points interferes with polygonization step building roof top Therefore, it advantageous detect remove that represent both LiDAR imagery. paper propose a two-step method detection consisting segmentation followed by classification. done using simple region growing algorithm weighted features LiDAR, as height, texture map, height variation, normal vector estimates. weights determined learning on random walks. classification support machines (SVM), allowing us control misclassification rate. overall formulated binary problem, results presented receiver operating characteristic curves shown validate our