Object Recognition using the Kinect.
Independent thesis Advanced level (professional degree), 20 credits / 30 HE creditsStudent thesis
Object recognition using the Kinect
The analysis of the environment is of great importance for autonomous robots because their behavior in the navigation depends on this. Depth distance from objects into the scene then represents a very useful data. As this sensor type previously existed only in very expensive devices, there is a lack of software and algorithms that exploit it comparing to the 2D space where usually a single camera is used without the information of the depth.
Since the Kinect has been introduced into the market in 2010 at a really cheap price for this kind of device, a lot of research is involving this field. Inside the Robot Operating System (ROS) software the Point Cloud Library (PCL) has been integrated. On this library some state-of-the-art algorithms have been created. We will either use the ROS and the PCL library, evaluating the results and introducing some improvements and hints for future work. We will introduce also some state-of-the-art algorithms in order to apply the conversion from Computer-Aided Design (CAD) into Point Cloud Data (PCD) models which represent the format used by the PCL.
Place, publisher, year, edition, pages
Trita-CSC-E, ISSN 1653-5715 ; 2011:115
IdentifiersURN: urn:nbn:se:kth:diva-130762OAI: oai:DiVA.org:kth-130762DiVA: diva2:654209
Master of Science in Engineering - Information and Communication Technology