Depth Image Based 3D Body Recognition And Surface Reconstruction In Real-time Augmented Reality

Three-dimensional reconstruction technique is an integral part in the field of augmented reality. It have a very wide application prospects in many field like movies, games, human-computer interaction, motion analysis, medical diagnosis, etc. And 3D body surface reconstruction is an important branch of the three-dimensional reconstruction techniques, it has been very widely used in many fields.For augmented reality technology can’t meet the outstanding issues of real-time, and the importance of three-dimensional human body surface reconstruction technique for augmented reality human-computer interaction, we presents an innovative real-time three-dimensional reconstruction algorithm based on body surface depth of the image in augmented reality only based on depth sensor for input data.On the part of pretreatment of depth image, we analysis the data stream of Kinect firstly, then use a specific function to get the depth data stream needed to reconstruct the body surface. Secondly, we use some common different smoothing filter for smoothing the depth information and comparing the different results of smoothing filter. At last, in order to achieve the purpose of real-time, we accelerated the median filtering processing.On the part of three-dimensional human body surface reconstruction, by way of depth information interception in dual threshold, we can cut out unnecessary data collection and compute the 3D point cloud data at the same time. Further, we transfer the point clouds to surface through triangulation line by line. Lastly, calculating normal vector of point cloud can help to render the surface and smoothing filtering can smooth the surface.Lastly, we set sophisticated experiments for time complexity and comparing with existed algorithms. We proved that our solution has many advantages against common depth based reconstruction methods.