Three-Dimension Digitization Of Human Foot Based On Computer Stereo Vision Combined With Kinect Sensor

With the continuous development of computer vision theory,virtual reality technology has been applied to various fields of life,including aerospace,virtual and reality,reverse engineering and medical fields.Three dimensional reconstruction of objects has become one of the hottest research fields.The research of the foot type digitization is a reverse engineering,and the reverse engineering is an important branch of the computer graphics field.This paper studies the present situation of foot type digital measuring equipment,because of the existing high price and long scanning time,requirements for the operator,high degree of automation is not higher shortcomings,therefore it needs a rapid and simple method,high precision of human foot 3D digital.Microsoft Corp launched a somatosensory peripheral Kinect,which built a depth camera,the use of active infrared structured light to get the object's three-dimensional point cloud data,not affected by the texture characteristics of objects.In this paper,a three dimensional digital model of foot is proposed,which combines the Kinect sensor and the computer stereo vision.The research contents of this paper are as follows:(1)The construction of Kinect sensor and binocular camera system.The main contents of this part are as follows: building the 3D measurement system.The calibration method of Zhang is used to calibrate the 3D measurement system.First of all,the epipolar line correction of the acquired image and depth information is obtained.The stereo matching is used to obtain the disparity map,and then the 3D point cloud of the object is obtained.The depth information image acquired by the Kinect sensor is processed by filtering,denoising,background filtering and so on.(2)Feature extraction and matching based on Harris-Laplace-SIFT algorithm.Firstly,the RGB images of different views are represented by multi-scale,and the feature points are extracted by Harris algorithm.The extracted feature points are described by SIFT descriptor and the feature vectors of 128 dimensions are generated.It is found that the nearest neighbor points and the nearest neighbor points,such as the ratio between them,are less than a certainthreshold value.(3)Splicing of point clouds in ICP algorithm.Since the center of the different coordinate system is not unified,this paper proposes an initial registration algorithm based on feature matching.The method uses the point cloud of the extracted feature points as the initial registration points to obtain the optimal parameters.And then,the 3D point cloud of the human foot is formed.The experimental results show that the proposed algorithm can reduce the computational complexity and improve the registration efficiency.First of all,the 3D data points of the fused 3D data points are processed by triangle mesh.The software of Geomagic is used to reconstruct the contours and surfaces.Secondly,to optimize the processing of the surface of the grid structure,and then NURBS(Non-Uniform Rational B-Spline)surface fitting,rendering a realistic three-dimensional foot model.Finally,the feature parameters of human foot are extracted.