Low-Cost Stereo Vision: A Single-Camera Approach For Precise Robotic Perception

In industrial robotics, the perception and measurement of the robot’s environment plays an increasingly important role in the development of flexible and adaptive robot systems. Available systems such as 3D cameras, laser sensors or stereo cameras for determining 3D coordinates are usually very expensive to purchase, but deliver good accuracy values in the millimetre or even sub-millimetre range. The aim of this paper is therefore to present a method for stereoscopy using a single inexpensive camera that can achieve similar accuracy to existing technologies. The camera is mounted on the flange of a robot arm, which enables images of a scene to be taken from different positions and orientations, and depth data to be determined by means of triangulation. In addition to measuring individual points with this type of stereoscopy, it is also conceivable in later development stages of the methodology to generate entire 3D scans of objects due to the flexibility introduced by the robot.