Book Description
This project is based on the recent advances of 3D depth cameras, one being the Microsoft Kinect sensors. Nowadays, this technology enables to create really accurate 3D figures and scenes simplifying difficult tasks in Robotics and embedded systems. The purpose of this project consists of using both the color data and this depth sensing technology to reconstruct and recognize objects in a scene. The data from the Kinect sensor is processed first to eliminate the planes of the scene to get a better perception of the elements we may find, and consequently, to obtain relevant features of every single object for the creation of a model for every different class. The results show the performance of the system, by testing a database of different elements against the training of some specific objects previously selected: Bowls, pillows and monitors. We can conclude that the classification using the color information is more accurate than using 3D data, even though in both cases the results are quite satisfactory.