ToF 3D Vision Algorithms in C++ for Robotic Applications

The present work focuses on learning how to use the hardware and software of a Time-of-Flight camera and then creating C++ programs that try to lay the basis for future ones that would be implemented in bigger applications that are able to tackle and solve the most common problems related to 3D machine vision. All the code and programs present in the thesis are born and developed with the key idea in mind that they can provide a foundation for future work as the starting point for more complex projects related to 3D computer vision and thus be implemented in applications such as bin-picking, autonomous driving, quality inspection and so on. The C++ programs developed in this thesis, along with smaller questions, address four main problems in the field of computer vision: Registration, CAD model creation, Segmentation and Object Recognition. Registration consists of aligning various 3D point clouds taken from different points of view of the same object so as to reconstruct a full 3D model of the object. CAD model creation is usually the step following Registration and it consists in creating a 3D mesh of the registered object. This mesh can be exported in AutoCAD or any other 3D CAD software to perform further studies and analysis on the object. Segmentation consists of dividing point clouds in multiple parts and pieces, which can be then studied independently. Finally, the last topic studied and developed in the thesis is Object Recognition, which consists in the ability to locate and recognize objects from a 3D point cloud.