Research On Indoor Navigation Method Of Micro Rotor UAV Based On 3D Scene

Micro rotor unmanned aerial vehicle(UAV)has the advantages of small volume,low cost,easy to use,can replace human to do indoor target search,NBC information detection and other tasks.Therefore it has great potential applications in promising counter-terrorism,investigation,emergency rescue and other fields.It is a great challenge to realize indoor autonomous navigation for UAV because that the complex indoor environment,the space structure,the obstacle and the personnel walking will all affect the UAV’s flight.Focusing on the indoor navigation of UAV,we study the modeling method of indoor 3D scene,the extraction method of UAV flight area and the realtime path planning technology in this paper.The main work and contributions are as follows:(1)The method of indoor 3D scene modeling is studied and a spatial semantic structure for indoor navigation of UAV is proposed,providing support for UAV flight area selection and path planning.The characteristics of indoor space are described for unmanned indoor navigation;According to the characteristics of indoor space,the semantic classification of indoor elements is carried out and the semantic structure of indoor space navigation is defined;The 3D geometric model of the experimental scene is established by using the method of manual modeling.Finally,the semantic information and navigation network information are added into the model and expressed based on CityGML and IndoorGML.(2)The method to extract and visually express navigable space of UAV is sutdied.The 3D geometrical model is voxelized based on the octree structure.The intersection is calculated between voxel space and geometrical location of the building model.Then the occupied space of voxels and non-occupied space of voxels are able to be obtained.The OctoMap is introduced as a probabilistic,flexible and compact 3D map representation for robotic systems.The navigable space of UAV is visually expressed and analysed based on the Octo Map.Experiments prove that this method is feasible,and can provide effective technical support for the UAV path planning.(3)A 3D path planning method for UAV flying in the indoor environment is put forward.The surfaces are segmented by using flood-fill algorithm.The geometrical features of the model are used to make a distinction between floor,stairs and obstacles.All horizontal surfaces are now labelled as one of the three classes: floor,stairs or obstacle.These labels are propagated upwards to also label the empty space above the horizontal surfaces.The distance field is computed using CDT.The distance field is segmented into cells using the watershed transformation.The graph is derived from a cell decomposition of the navigable space.This cell decomposition is realized by performing a watershed transformation on a distance field of the input model.The shortest path between the two newly inserted nodes is found by running the A* algorithm.It can be concluded that current methodology works well for flying actors in the 3D path planning indoors.