Augmented Privacy with Virtual Humans

摘要

Visual privacy is a sensitive subject because it literally deals with human private parts. It presents a bold challenge to the field of Computer Science. The goal of this study is to build a virtual human model for designing and evaluating visual privacy technologies before a security system is built. Given the available databases of anthropological models from CAESAR, 3D scanners and the physical parameters of human imaging systems, we simulate the scanning imagery data with the High Frequency Structure Simulator (HFSS). The proportion and template matching algorithms have been developed to find the human surface features from 3D scanning data. The concealed object detection algorithms are developed according to the wave intensity and surface characteristics. Then the privacy-aware rendering methods are evaluated by usability studies. This forward-thinking approach intends to transform the development of visual privacy technologies from device-specific and proprietary to device-independent and open source.rn It also advances privacy research from an ad-hoc problem-solving process to a systematic design process, enabling multi-disciplinary innovations in digital human modeling, computer vision, information visualization, and computational aesthetics.rnThe results of this study can be used in the privacy-aware imaging systems in airports and medical systems. They can also benefit the custom-fit products that are designed from personal 3D scanning data. Furthermore, our results can be used in the reconstruction of objects in digital archeology and medical imaging technologies such as virtual colonoscopy.