In order to realize three-dimensional reconstruction of canopy at different growth stages of fruit trees, 3D data acquisition methods and canopy reconstruction methods were studied. Based on the analysis of morphological and structural changes in fruit phonological phase, and integrating the advantages of different data acquisition techniques, the data acquisition method of fruit tree morphological structure based on multi-source is proposed. In the dormant period, the canopy skeleton is extracted based on point cloud data; in the leaf curtain stage, a new artificial coding method of canopy structure is constructed, and the data of new shoots and leafs is obtained efficiently; and organ template data is obtained synchronously, and the organ template library is constructed. Then, a multi-source data fusion modeling method is proposed to reconstruct the three-dimensional canopy of fruit trees at different growth stages. And the feasibility of the method is verified by 12 year old open central leader system apple trees, the results show that compared with the manual data acquisition method, the method improves the efficiency by more than 5 times, and the error rate is less than 6%. It provides a feasible scheme for the continuous data acquisition and canopy 3D reconstruction of fruit trees, so as to provide technical support for virtual modeling, scientific calculation and experimental simulations.