Monocular Cloud Map Generation for Intelligent Navigation

The existing Intelligent Navigation systems for Robotic operations suffer from high-cost requirements of the camera modules or the sensory tracking modules used in the Simultaneous Localization and Mapping (SLAM) technique. The sensor in such systems needs to be replaced with high-end cameras which can provide depth information as well, which further adds to the cost of the system. The depth information is necessary to build 3D maps of the environment for further navigation requirements. This makes it ineffective for everyday in-home applications. To solve this, a system that can use available smartphones to perform sensing operations required for Oriented FAST and Rotated BRIEF-SLAM (ORB SLAM) is proposed. The proposed system performs object detection of common household objects using the YOLO algorithm along with CNN networks.Utilizing a smartphone camera, a point cloud map has been designed using the ORB SLAM algorithm with re-localization, loop closing, and map reuse features. Tested with indoor sequences, the integrated Robot Operating System (ROS) provides exemplary performance in real-time. The targets could be detected using ORB SLAM, and point clouds that display and map the surrounding spaces and explore the unknown environment were created. This system eliminates the need for expensive 3D cameras or depth sensors by providing a cheaper alternative using any monocular camera and performing operations like point cloud mapping, depth mapping, and object detection in real-time.