Virtual Human Generation In Real Time Based On Body’s Basic Information

Using cheap camera for the 3D Human Body Reconstruction has so much value and importance. Currently, works in Body Reconstruction with raw scan data cannot run in real-time. Real-time Body Model tracking methods are all based on templates, requiring accurate input data and templates, thus cannot be implemented in low-cost devices. There will be more applications in inexpensive camera to do real-time Human Body Reconstruction, thus it requires more methods to do interaction between computer and Virtual Human Body which contains basic body information.Human Body Model Reconstruction is a key technology in Human-Computer Interaction and Virtual Try-on, but it has not been implemented on low-cost devices, for this situation we proposed a novel algorithm to use one Kinect camera to reconstruct the human body based on basic human body information in real-time. Our method use basic body information according to feature-circles to reconstruct Human Body Model by generating mesh between two feature-circles. Feature-circle is the intersection of body and the plane which is decided by a joint in a bone and the bone-vector, the radius of feature-circle can describe thin-fat information of this part of the body, and the distance between two feature-circles can describe the height information of the body part. Results of this paper indicate that our feature-circle-based Human Body Reconstruction method has less vertices and good real-time performance, thus can be used in real-time applications which has less demands in model accuracy.