Study on the three-dimensional (3D) canopy reconstruction of fruit trees plays a fundamental role in the calculation of canopy illumination distribution and the realization of automatic pruning. According to the characteristics of the leafy tree point cloud, an apple tree leaf points clustering method based on dynamic-K-threshold and a growth parameters extraction method were proposed. Firstly, terrestrial laser scanner Trimble TX8 was chosen to obtain dense point cloud of apple tree canopy from different viewpoints, and then multistation point cloud registration, outlier removal and point cloud simplification were accomplished, so as to reduce the influence of discrete points on calculation results of spatial characteristic parameters. Secondly, intercepting one branch randomly, synthesizing LCCP algorithm and improved K-means algorithm to construct the leaf points clustering method based on dynamic-K-threshold. Thirdly, as the input data, the leaf point cloud was used to construct the covariance matrix based on the PCA to calculate the fitting plane normal vector. Extracting boundary points, the parameters of width and length were obtained by calculating the position relation between boundaries and centroid. The results showed that compared with traditional clustering methods, the proposed dynamic-K-threshold method can accurately segment single leaf points without branch point losses, which ensured the integrity and thoroughness of the clustering results. The extracted parameters based on real 3D spatial information can guarantee the accuracy to a certain extent, which provided basic technical support for evaluation of illumination distribution calculation and automatic pruning.