Indoor Localization Through the Integration of RGB and Depth Data from Microsoft Kinect Sensor

L. Magasweran, J. Huang

Abstract:	Mobile robots operating indoors often must be capable of navigating without access to GNSS (Global Navigation Satellite Systems) data. Determining the indoor location is one of the key functions of mobile robots before they can operate properly. Existing indoor localization techniques, such as WiFi beacons and SLAM (Simultaneous Localization and Mapping), are either susceptible to tampering, environmental noise, or are costly. Alternatively, existing permanent indoor landmarks such as doors, windows and light fixtures can be identified through image processing techniques and used to determine robot location based on known landmark locations. In this research, a localization algorithm that combines visual and range data using the Microsoft Kinect sensor, libfreenect driver, and OpenCV computer vision software is proposed. The Microsoft Kinect sensor is a low-cost device that combines multiple sensors to provide both depth and RGB video data via USB. In addition to using commercial off-the-shelf hardware, this algorithm leverages the OpenCV (Open Source Computer Vision) library, which provides many robust and easy to use functions. The algorithm proposed in this paper exploits a Kinect depth sensor "limitation" to reliably identify a room's features and from them triangulate the robot's position. The proposed algorithm first identifies landmarks by searching for quadrilaterals that exist in color data but absent from the depth data. The perspective transformation of these landmarks and the depth to their vertices are then used to estimate the robots relative position. Finally, with a priori knowledge of the fixed coordinates of those landmarks, the robot's absolute position with respect to the room frame can be estimated. This research will present the algorithm as well as preliminary results from experiments conducted in a controlled environment.
Published in:	Proceedings of the ION 2013 Pacific PNT Meeting April 23 - 25, 2013 Marriott Waikiki Beach Resort & Spa Honolulu, Hawaii
Pages:	510 - 518
Cite this article:	Magasweran, L., Huang, J., "Indoor Localization Through the Integration of RGB and Depth Data from Microsoft Kinect Sensor," Proceedings of the ION 2013 Pacific PNT Meeting, Honolulu, Hawaii, April 2013, pp. 510-518.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In