In this paper, a Legorization method which can reconstruct LEGO model with complex internal and external structures from 3D color printing trajectory is proposed. Different from voxelization methods, by combining advanced adaptive slicing algorithm with building “high-resolution” regions with thin plates, the reconstruction accuracy of initial LEGO units can be guaranteed. Furthermore, the tree structure is employed for automatically generating support structures which can be converted into LEGO support structures. By adopting split assembly appropriately and implementing combination of these parts accurately, the reducing supporting structures can be further simplified. In order to optimize the Legorization scheme, a machine learning method is used to guarantee the quality and efficiency of the reconstruction work. Complex LEGO models are provided to demonstrate the effectiveness of the proposed method.