Semantic Mapping Of Indoor Scenes

In the autonomous navigation and path planning of aircraft and ground robots,the establishment of semantic map is very important.This paper focuses on the establishment of the indoor semantic map,including the definition of semantic map,the construction of spatial 3D map and the target detection of indoor scenes,with the background of the eighth generation task of the air robot competition.And semantic map fusion and so on.The main contents of this paper are as follows:First,the inter frame alignment method based on sparse feature point information matching and dense point cloud information error minimization is given,and each frame position adjustment method based on global optimization is given.The method of point cloud splicing and optimization and dense point cloud map generation are given,and the grid based on point cloud map is generated to occupy map generation.Method.Secondly,the target location method based on Selective Search algorithm is given,and the target identification method based on support vector machine is given,and the target region segmentation method based on the region growth of the depth image is proposed,and the semantic information of the target is generated.Then,a semantic fusion method based on maximum likelihood estimation is proposed,which generates semantic tags from semantic sequences.The semantic tags are mapped to the map to generate the final semantic map.Finally,the algorithm is given and the algorithm is implemented.The feasibility and effectiveness of the algorithm are illustrated by simulation and analysis.