Font Size: a A A

Research On 3D Human Body Reconstruction And Fusion Algorithms For Virtual Reality Environment

Posted on:2020-05-22Degree:MasterType:Thesis
Country:ChinaCandidate:R H YanFull Text:PDF
GTID:2428330623963634Subject:Computer technology
Abstract/Summary:PDF Full Text Request
Virtual reality(VR)has important application value in reality.In many areas of research and commerce,virtual reality is a hot spot.The boundary between virtual world and real world is interaction.The interaction in VR environment can make the information between virtual world and real world interoperable.Reconstruction can show the information of user objects in the real world in the virtual world.However,the current reconstruction technology has the shortcomings of high equipment requirements and slow speed.Moreover,due to the interference of VR devices,the acquired images often have noise.To solve these problems.In this paper,human interaction technology in VR environment is studied from three aspects: 3D face reconstruction,3D human body reconstruction and motion capture.In this paper,besides 3D face reconstruction,there are two pre-steps of face detection and face restoration.Inspired by the principle of pinhole camera,our 3D face reconstruction technology uses the labeled 2D-3D paired objects as training set.To obtain the transformation relationship between [2D,1]~T and the corresponding objects in the three-dimensional world.The main innovations of this paper include the following aspects:1.A face recognition method based on Haar feature and skin discrimination is proposed to improve the recognition rate.2.A deep convolution neural network is proposed to realize face restoration.In addition,the global and local discriminant networks are used to test the repair results to extract better repair results.3.A series of regression methods are proposed to calculate the position coordinates of facial features,and then the corresponding points of the corresponding three-dimensional model are matched as the training set.To achieve faster 3D face reconstruction based on single image.4.A motion capture method using elliptical skeleton instead of linear skeleton is proposed.Because the ellipse can represent more rigid space than the line segment.So using elliptical skeleton can express more information.5.A three-dimensional human body reconstruction method based on box cutting fusion method is proposed.The object's body is divided into several parts.After segmentation,the local bodies are enclosed in boxes.Finally,the three-dimensional model is obtained by fitting and transforming the skin.
Keywords/Search Tags:Virtual reality, Face reconstruction, 3D human body reconstruction, Motion capture
PDF Full Text Request
Related items