Abstract: Aiming at the problems of poor accuracy and lack of overall structural understanding of the automatic three-dimensional reconstruction method using traditional monocular camera,a full-automatic indoor three-dimensional layout reconstruction system combining visual inertia odometer(VIO)and motion from structure(SfM)is proposed.Firstly,the visual odometer is used to obtain the key frame image sequence and the corresponding spatial position and pose,and the SfM algorithm is used to calculate the precise camera pose.Then,the multi-image stereo geometry algorithm is used to generate the high-quality dense point cloud.Finally,based on the hypothesis of Manhattan world,a rule-based bottom-up layout reconstruction method is designed for typical interior scenes of modern buildings,so as to obtain the final outline layout of rooms.The experimental data of the Zhejiang University CAD&CG laboratory scene scan data set and the synthetic dense point cloud data set were used as experimental data,and experiments were carried out under the Ubuntu 16.04 and PCL 1.9 environment.The results show that the proposed method has high tolerance to 3D point cloud noise and can effectively reconstruct the 3D outer contour layout of indoor scenes.