Efficient Transmission and Rendering of RGB-D Views

For the autonomous navigation of the robots in unknown environments, generation of environmental maps and 3D scene reconstruction play a significant role. Simultaneous localization and mapping (SLAM) helps the robots to perceive, plan and navigate autonomously whereas scene reconstruction helps the human supervisors to understand the scene and act accordingly during joint activities with the robots. For successful completion of these joint activities, a detailed understanding of the environment is required for human and robots to interact with each other. Generally, the robots are equipped with multiple sensors and acquire a large amount of data which is challenging to handle. In this paper we propose an efficient 3D scene reconstruction approach for such scenarios using vision and graphics based techniques. This approach can be applied to indoor, outdoor, small and large scale environments. The ultimate goal of this paper is to apply this system to joint rescue operations executed by human and robot teams by reducing a large amount of point cloud data to a smaller amount without compromising on the visual quality of the scene. From thorough experimentation, we show that the proposed system is memory and time efficient and capable to run on the processing unit mounted on the autonomous vehicle. For experimentation purposes, we use standard RGB-D benchmark dataset.