3D Modelling of Dynamic Objects and Complex Scenes
Abstract: 3D digital representation of real-world shapes and scenes is essential for a variety of applications ranging from virtual/augmented reality to geometric modelling for 3D printing. Due to unavoidable occlusion and sensor limitations, multiple scans are often needed, which are then registered to form complete shapes. Despite great effort, this is still a challenging problem, especially when low-cost colour and depth (RGB-D) sensors are used, and objects are deforming non-rigidly. My talk will overview some of our recent effort in addressing this problem, including identifying correspondences efficiently and robustly using local geodesics and diffusion pruning, improving robustness and accuracy of non-rigid registration using sparse priors, and exploiting knowledge in a shape collection to effectively register a template model to a single noisy scan captured using a low-quality depth camera. Short bio: Dr Yu-Kun Lai is a Senior Lecturer at Visual Computing group, School of Computer Science & Informatics, Cardiff University, UK. He obtained his bachelor’s and PhD degrees from Tsinghua University in 2003 and 2008, respectively. His PhD thesis received the National Excellent Doctoral Dissertation of China Award. He joined Cardiff University in 2009 as a lecturer. He has been working on broad areas of Visual Computing, including Computer Graphics, Geometric Processing, Image Processing and Computer Vision. His research has been supported by EPSRC, Royal Society and industry. He has published 50 papers in world-class journals, including ACM Trans. Graphics, IEEE Trans. Visualization and Computer Graphics, IEEE Trans. Image Processing etc. He is on the editorial board of The Visual Computer, conference co-chair of SGP 2014 and CVM 2016, and on the programme committee of a number of major international conferences.