Modified Indoor Navigation Mapping based on VSLAM

Since the 2D maps built by traditional method are limited to describe single environment plane, the VSLAM-based indoor 2D map building method is proposed. The nonlinear optimization problem about camera poses is constructed based on VSLAM, and this problem is solved by Gauss-Newton method. Subsequently, the point cloud map of indoor environment is built. Then, based upon the RANSAC algorithm, the equation of ground plane is obtained. The point cloud of robots' passage area is projected onto the ground to form the grid map, and the indoor maps can be built by updating the grid occupied state based on Bayesian method. Experiments show that, compared with the 2D maps built by the traditional method, the maps built by our method contain the 3D structure information of environment. It helps robots avoid the suspended objects and the low objects successfully in the path planning experiment.