Reconstruction of scenes using a hand-held range imaging càmera

Abstract

The current nal degree project covers the development of a C++ program to reconstruct 3D scenes from a stream of incoming RGBD images from the Kinect for Windows v2 sensor using the PCL library and the libfreenect2 open source Kinect drivers. A background analysis is performed to analyze state of the art registration algorithms and similar registration applications which are already distributed. All the computing steps are explained and detailed, from the data acquisition through the coarse and ne alignment steps to the nal surface reconstruction algorithm. Several algorithms are discussed for each step, giving reasons for the nal algorithm choice and the parameters set. Finally, a brief discussion on the results and problems encountered during the development of the project serves as the basis of a proposal of improvements for future work on the subject. The project budget and the environmental impact are analyzed before the nal conclusion chapter.