Research On3D Body Reconstruction And Measurement Techniques Based On Kinect

EMTM and personalized production of the costumes have been the developments tend for the costume industry. The appearance of the non-contact3D body scanner has made it to be true. Presently, non-contact3D body scanning technology is not very advanced in our country. Non-contact3D body scanner from foreign country is so large and expensive that many costume manufacturers can not afford it. Meanwhile, using traditional3D body scanner to obtain the characteristic data of the body surface is very complicated. Therefore, exploring the small volume, high precision and low cost3D body scanner has comprehensive value in research and prospect for application. In this thesis, a scheme of3D body scanning system based on Microsoft Kinect has been proposed. The main contributions of this thesis can be summarized as follows:Using Microsoft Kinect to obtain the depth body data and get the depth image. In this thesis, the depth scanning principle of the Microsoft Kinect has been researched. First, depth image is taken by Kinect. According to the characteristics of the depth frame, the information of the model will be gotten. Then, the threshold method used in digital image processing is introduced into extraction of depth image. Depth dual threshold method is proposed. The model’s depth image can be extracted entirely by this method.Point cloud data processing. First, the depth data obtained by Kinect are transformed into the form of3D point cloud to store and visualize. And then, make rejections, filtering, and simplification for point cloud. In the process of simplification, we take the advantage of the minimum distance method and the angular deviation method, introduce the parameter of curvature and put forward an improved self-adapting method of simplification.Splicing different part of the body model point cloud and rebuilding the body model. There are various methods of splicing the point cloud and rebuilding the model. After analyzing many popular methods, the body model point cloud data taken from different perspectives by Kinect are spliced by coordinate alignment method. And the body model is rebuilt by Delaunay algorithm.Preliminary study on human body feature size measurement. In this thesis, some characteristics of human body are defined and described. According to the characteristics of different parts from the actual body model, they can be identified. The body curve can be approximate as ellipse and the characteristics size of the body can be calculated3D body scanning system based on Microsoft Kinect has the advantage of small volume, convenience and low cost, and it has a broad perspective in manufacture.