A Novel Deformable Model for Urban Vegetation Detection Using LiDAR Data

This paper presents a new approach to creating variational level set model for vegetation detection combining 3D irregular point clouds and aerial image simultaneously acquired by LiDAR light scanning and imaging device. Firstly, a fundamental statistical level set framework is built which integrates texture information to improve the quality of vegetation detection. Then, several derived products directly or indirectly from LiDAR raw point cloud data, like DTM(digital terrain model,) nDSM(normalized digital surface model) and local roughness capable of describing 3D texture feature of vegetation, are used to construct a novel energy term in relation to height and roughness of non-terrain objects, in order to make up the disadvantages caused by insufficient information only from remote sensing image. This model can well fuse spectral feature, height and roughness information of objects from different sensors. Experiments on pairs of LiDAR Aerial image and 3D point cloud data are carried out, and conclusions can be drawn that our model can effectively separate various vegetation categories including grass and tree in urban area from other land covers, including buildings, noises, ground etc., and alleviate various influences caused by occlusions or spectral inhomogeneity.