Abstract:The fine structure inside the canopy of apple tree determines the light distribution and is one of the important factors affecting the quality and yield of apples. Aiming to rapidly generate apple tree canopy 3D reconstruction model to study the illumination distribution in it, a point-leaf model reconstruction method of apple canopy based on point cloud data was proposed. A Trimble TX8 was used as a data acquisition device to obtain 3D point clouds of the canopy in the flowering, leaf growth stage and stable growth stages of apple tree as the research object. Firstly, a leaf spatial location (LSL) extraction approach using densitybased spatial clustering of applications with noise (DBSCAN) and layers Kmeans and median methods was proposed. In the LSL extraction progress, the DBSCAN clustering method based on adaptive parameters was used to extract the point cloud of singleleaf branches. And the same point cloud was sliced into layers, and the Kmeans and median method was used to fit the branch center line. And then, the Euclidean distance of each point between the single blade and the center line was determined, and the point with the smallest Euclidean distance (ELD) was the LSL point. Field experiments showed that the method proposed was suitable for LSL extraction during leaf growth and stable growth stage. Through comparative analysis between the automatical algorithm and artificial use of Realworks software extraction point showed that the average ELD between these two methods was 1.41mm. Secondly, the whole apple tree canopy LSL was extracted by canopy voxelization, and the point model was constructed. Using the leaf templete to splice the LSL, the point-leaf model was formed. The light interception calculation (STAR value) in the VegeSTAR light model showed that the 3D reconstruction efficiency was improved greatly. Therefore, the point-leaf model reconstruction method was effective.