An Extensible 6DOF Vision Framework for Accurate Location and Pose Estimation in a Scene Graph Context

RoadNarrows proposes a novel architecture to be implemented in a 6DOF vision system with applications in autonomous robotics. The system will report objects of interest along with location and pose information to a supervising computer. This system will be built around an advanced 3D vision platform that can deliver multiple video streams and 3D spatial measurements. The sensors will deliver high-resolution real-time spatial data which will be incorporated into an innovative framework for creating a scene graph approximation of the environment. Each set of measurements received from the sensor will be used to refine this scene graph model. The scene graph will contain important information about all objects of interest in the environment such as: estimated location, pose, and velocity. This approach, which builds and refines a coherent model of the environment, has significant advantages over simpler vision systems that analyze data frame by frame.