Abstract: Four representative street trees (magnolia grandiflora, sapindus mukorossi, platanus acerifolia, cinnamomum camphora) in Haining city, Zhejiang province are selected as research objects. Combined with unmanned aerial vehicle(UAV) images and 3D laser scanning data, the point cloud splicing, filtering, noise reduction and editing are completed by using ContextCapture and LiDAR 360 software, and the point cloud fine matching is achieved by iterative closest point algorithm. Complete multi-platform point cloud data fusion, obtain digital surface model and digital elevation model and make canopy height model on this basis; Watershed segmentation algorithm is used to segment CHM of different street tree species, and local maximum method is integrated to extract parameters of single tree height and crown width. The data show that the proposed method can achieve high precision of single tree segmentation, and the extraction effect of tree height and crown width is good, which meets the requirements of street tree geometric parameter investigation.

Key words: street tree, UAV image, 3D laser scanning, canopy height model, watershed segmentation algorithm, tree height, crown width