In order to guide the pruning, flower thinning and harvesting of fruit trees in orchard, a novel vision system which combines a color-camera system with a photo mixing detector (PMD)-camera is constructed. For the three-dimensional coordinate information of target scene acquired by the PMD camera, effective point cloud combining PMD amplitude image with density-based spatial clustering of applications with noise (DBSCAN) algorithm is extracted. Multi-source information fusion is completed with the result of image registration in the previous studies. Primary component analysis algorithm (PCA) is used to get the initial state of the point cloud at different locations, which is called prealignment. Accurate splice between two point clouds is realized by the iterative closest point (ICP) algorithm based on the least square method to get the optimal matching. Coordinate transformations are obtained by singular value decomposition (SVD) after prealigment and accurate splice. Several groups of experiments are used for verification, which show the average error of multi-view point cloud splicing reaches 2.62 cm and can better make up full three dimensional display of apple tree canopy without missing data than a single angle shot.