Abstract: Aiming at the difficulty of accurate mapping in the process of single view 3D reconstruction of complex topological objects, a single view 3D reconstruction method combining convolution Catmull-Rom spline free-form deformation (CR-FFD) and offset Transformer capsule network is proposed in this paper. Firstly, Catmull-Rom spline basis function was used to interpolate the control points of the point cloud model to maintain the consistency of the local topology of the point cloud model deformation. Then, the least square inverse solution method based on convolutional neural network is used to accelerate the solution process through nonlinear parameter mapping. Finally, a Transformer encoder module based on offset attention is designed and embedded into the capsule network to enhance the local feature expression ability to capture the fine-grained features of the shape of the point cloud. Experiments on ShapeNet dataset show that the average values of EMD and CD are 3.84 and 3.71, respectively, and the average values of EMD and CD in Pix3D dataset are 5.51 and 5.39, respectively. Compared with the existing single-view point cloud 3D reconstruction methods, the proposed method effectively improves the 3D reconstruction results of single-view point cloud, and can maintain the consistency of reconstruction from different perspectives.