REAL-TIME SCENE FLOW ESTIMATION AND OBSTACLE AVOIDANCE UNDER DIM-LIGHTCONDITION USING TIME OF FLIGHT (TOF) SENSOR
Sobers LX Francis, Sreenatha Gopalarao Anavatti, Matthew Garratt · Journal of Critical Reviews · 2020
The paper introduces the novel method for the estimation of dense scene flow using RGBD camera`s range data. These calculated flow vectors are used to discover the obstacles, so that the mobile robot`s intelligence for the path planning module can be improved. These mobile robots need to be clever enough to discover their workspace and correctly keep away from obstacles. In a cluttered area, to avoid obstacles, the AV requires complete awareness of the surrounding. Hence, the speed and direction of dynamic obstacle are detected by comparing successive range frames which are obtained from single vision sensor. The current research deals in regard with the extraction of the 3D flow vectors from the differential scene flow technique. The Gradient Vector Field (GVF) technique is used with scene drift to decide changes inside the pixels in order to recognise all obstacles in three directions. Our goal is to develop a real-time scene-flow based obstacle avoidance technique that relies on a single sensor perception system. Experimental result is shown to demonstrate the efficacy of the technique under dim night-time lighting condition using 3D sensor.