3D Face Recognition: Image Acquisition
F.S. Fikke, B.K. Gardiner · 2013
The ability to reliantly identify people is an ever growing desire. 3D imaging devices provide major new possibilities in this field. This thesis presents a method which uses a Microsoft Kinect to acquire a depth map of a persons face and transforming this into a point cloud that can be used for further processing into a 3D model of a subjects face. The thesis starts by giving a description of the requirements our system needs to abide by to achieve the desired output. Since this output will undergo further processing, it is vital our output format matches the pre-defined format. After that the details of the Kinect will be discussed, including the technical architecture, Kinect performance, the method used by the Kinect to acquire depth data and the data structure it provides. Chapter 4 goes into detail on how this data is processed to extract the necessary information. Several filtering methods are introduced to forgo unwanted characteristics in the data. Finally in Chapter 5 we use the previous methods by applying them to real-world data and investigate the differences.