The Evaluation of Different Approaches Towards Using Kinect Sensor as a Lasor Scanner

Abstract

Transparent objects are easy to recognize with the naked eye, but due to the fact that infrared radiation is travelling through transparent objects, they are not sensed by robots using infrared technology. We propose a workaround to this problem, which improves Simultaneous Localization and Mapping (SLAM) performance. By performing more detailed scans with the Kinect sensor, we are able to find frames around the glass walls, and by that detect them as if they were solid walls. The proposed method is evaluated using the Microsoft Kinect sensor mounted on a Turtlebot robot. Our approach is continuously improved by using the black box software testing method, and we have accomplished good reliability in both the software simulator and the real robot. Results show that our approach gives approximately a 25% more realistic and correct recognition of transparent walls compared to built-in solution in environments that have such walls.