The objective of this project is to advance the state-of-the-art in acquiring and modeling dynamic non-rigid objects. The specific examples of non-rigid objects that the project is to focus on include: human faces, hands, soft tissues, cloths, and animals. The PI seeks to address the following two fundamental questions: (1) How can non-contact optical methods be used to measure dense 3D surface motion without physically modifying the appearance of the surface? (2) What physical and/or biological properties can be inferred from the acquired dense 3D motion data?
The research team addresses these two questions by two simple but general ideas, namely the space-time approach and data-driven models. The space-time approach builds upon space-time stereo, and enables accurate optical measurements of 3D surface motion, as well as automatic registration of shape sequences among different dynamic objects. The data-driven models are used for both material recognition and deformation-EMG correlation. The project has a wide range of scientific impacts, including generating data for 2D face alignment and 3D face recognition in biometrics, generating data for 3D face emotion recognition in human computer interaction, measuring human body deformation in biomechanics, modeling soft tissues for orthopedics and computer-aided surgery, and building virtual human models for entertainment and education. These scientific impacts translate into benefits to society, for example, by building more accurate biometric systems to secure our country, innovating surgery procedure to reduce health insurance cost, and creating 3D digital replicas of great teachers to make our education available anywhere, anytime, at a lower cost.
