Modelling and tracking the human from images are important topics in computer vision and artificial intelligence to give insights for understanding the human movements, with applications span from interactive media, biomechanical analysis, sports, wellness intelligence, future classroom instrumentation, and smart city with IoT. Together with augment reality, it could bring about another dimension of interactivity between the live and virtual.
Starting with 3D modelling of human, we proposed a method to reconstruct accurate 3D human surface skin and estimate its internal skeleton from several uncalibrated images. Facilitated by a regular surfaced 3D generic human model, our algorithm adapts it to fit the human in the images from several cameras and at the same time self-calibrates these cameras. The advantage of our approach is that since the customized model is an adaption through a generic model, the polygon and edge information are always consistent thus also the skin deformation function.
Next, we had proposed two methods for tracking of 3D human motion, one for monocular camera and the other for multiple cameras. Through a model-based analysis-by-synthesis approach we track the human pose by using numerical minimization as its kernel to search for the pose, in which we synthesized the 3D posture that minimizes the difference between the data of the synthesized movement and the real one.
The speaker will describe the current progress in human motion tracking and its potential research directions. Last but not least, he would highlight one of his recent completed industrial R&D projects titled Augmented Reality on Robotics Platform. This system incorporated inertial measurement unit (IMU) with simultaneous localization and mapping (SLAM), robot operating system (ROS) and Unity3D engine as the ground control station (GCS).
About the Speaker
QUAH Chee Kwang received his BEng from University of Leeds, England in 1996, MSc from University of Surrey, England in 1997, and PhD in Computer Engineering from Nanyang Technological University, Singapore in 2008. His research interests include computer vision, 3D human modelling, motion tracking and analysis, augmented reality, artificial intelligence, computer graphics, and applications to interactive media, smart city, education, defence and sports. He has many years of R&D working experience in these areas working at ST Engineering, Nanyang Technological University, Republic Polytechnics and DSO National Laboratories. Also, he was an invited researcher to INRIA, France during his PhD candidature.
He is a practice-track faculty candidate.
