Construction of 3D model of tree is a longterm research hotspot in botany, computer graphics, and architecture. And tree canopy branch reconstruction is an important component in the canopy dynamics analysis. The emergence of terrestrial laser scanners has accelerated this reconstruction process. To quickly reconstruct the canopy branch model, it is necessary to delete a large number of nonbranched interference point clouds. Taking the canopy of apple tree in the maturity growth stage as the research object, a method of colorbased sampling apple tree canopy trunk point cloud extraction was proposed. Firstly, the apple tree canopy color point cloud acquisition method was proposed. Trimble TX8 and coaxial panoramic camera were selected as the data acquisition device to acquire the apple canopy color point cloud data. Point clouds and color panoramic photos were matched in Realworks software, and color point clouds were get. Then, the color information R, G and B in the panoramic image was extracted. The adaptive segmentation threshold was established according to the distribution rules of R, G and B in the panoramic image branch area. Color point cloud data of the nonbranch part in the canopy was deleted according to the threshold. Finally, the 3D branch model was reconstructed in the Geomgic software. The process was followed by a series of operations, such as wrap, manifold creation, polygon editing, hole filling and smoothing. The experimental results of the apple tree branch extraction point cloud data showed that the point cloud deletion rate of this method was 75.74%. Compared with the artificial branch point cloud data extraction, the side branch accuracy rate was 93.34%, and the efficiency was improved by more than 200 times, shortening the threedimensional reconstruction time of canopy branches. In this way, the results of this study can provide a basis for studying the canopy structure analysis and the establishment of the branching dynamics model of the leafy apple tree.