Options
2018
Conference Paper
Titel
Dense 3D Environment Reconstruction with an RGB-D Camera for Mobile Robot
Abstract
In this paper, we present an environment reconstruction system to generate an indoor 3D map for mobile robots. Using an RGB-D sensor, the robot doesn't need the initial odometry. Furthermore, the system can be used for reconstruction of a 3D environment model by manually. We optimize our approach to reach 10Hz for the front-end and 1Hz for the back-end to fulfill the applications for the mobile robot. Our final goal is to develop the 3D SLAM systems which combine the Region Based Convolution Neural Network (MASK R-CNN) for creating a 2D semantic image and a 3D semantic map. The experimental results demonstrate some preliminary results for 3D reconstruction with an RGB-D camera for creating the point cloud map and the OctoMap for the mobile robot (Care-O-bot 4) in an indoor environment.