Evaluation of Kinect depth sensor for use in mobile robotics

Abstract

To autonomously navigate through unstructured environment mobile robots rely mainly on vision systems. Commonly used are stereo vision, laser and infrared sensors. This paper deals with an evaluation of the Kinect depth sensor and compares it with stereo vision. Also the depth resolution, the accuracy and the repeatability have been determined throughout several experiments based on the SDK of Code Laboratories (CL) for Windows 7. Our experimental results show that the Kinect has a maximum depth resolution of 1.3[mm/color value] and a maximum accuracy of 0.02%. Also, depth density has been assessed and compared to stereo vision images, concluding several benefits the Kinect brings to the field of mobile robotics. Finally, suggestions are being made for several applications for use in mobile robotics.