| In computer graphic and computer vision,3D reconstruction is to capture a real object and represent it as a mathematical model suitable for the representation and processing in computer.With the rapid development of 3D acquisition technology,3D data acquisition becomes more and more convenient,and 3D reconstruction is becoming a hot research topic in computer graphics.In this paper,we investigate the 3D hand modeling hand based on the Kinect,a depth sensing device developed by Microsoft.The device has the advantages of low cost,small size,simple operation,fast data acquisition,and so on.It receives more and more attentions in the field of 3D reconstruction and human-computer interaction in recent years.However,it is difficult to directly implement 3D reconstruction of real object,due to the low precision and noisy depth data acquired from Kinect.A hand modeling method based on Kinect is proposed in this paper.A rough hand model is obtained by scanning and reconstruction via Kinect,which has similar the size with real human hand.Then a template hand model is deformed to fit the rough model and finally a high quality hand model is obtained.The method consists following steps:first,two point clouds and their corresponding meshes of two sides of human hand are obtained via filtering and fusion based on KinectFusion algorithm.Then,the two point clouds are merger.Since the point clouds of both sides of human hand are not obtained at the same time,and human hands are non-rigid,a non-rigid registration algorithm based on thin plate spline is adopted to generate a rough hand model.Finally,the hand model is obtained by deforming an elaborate hand template to fit the rough hand model,which contains almost no geometric detail.Experiences show that this method can generate a high quality 3D hand model via acquisition and 3D reconstruction. |
