Research On Surface Feature Point Calibration Technique For Human 3D Reconstruction

The purpose of 3D human surface feature point calibration is to quickly and accurately locate the human body parts contaminated by radioactivity in order to facilitate emergency rescue work and protect the lives and health of staff and rescuers.In order to efficiently complete the accurate calibration of human surface feature points.In this thesis,3D reconstruction-based human dimension measurement combined with a scalable standard human model is used to achieve this.The main tasks are as follows:1)Human body point cloud data acquisition.Before starting to use Kinect to acquire depth image data,internal parameters are obtained by camera calibration.The imaging parameters obtained by camera calibration are used to convert the depth image acquired by the depth camera into a spatial 3D point cloud,and then extract the target human 3D point cloud by point cloud segmentation.Finally,after the down sampling process,the number of point clouds can be significantly reduced,thus greatly improving the accuracy and reliability of subsequent point cloud processing.2)Point cloud alignment.In order to avoid the problem of local optimization by directly using ICP algorithm to align the point clouds,this paper adopts a two-step alignment method to align the human point cloud data.Through the coarse alignment of the point clouds,the two point clouds get a better initial position,and then by comparing the experimental results of the two fine alignment algorithms,the VGICP algorithm with better results is chosen to complete the process,and finally the complete human point cloud model is obtained.3)Point cloud surface reconstruction.Firstly,the basic principles of the two algorithms,greedy projection triangulation and Poisson surface reconstruction,are introduced,and their application effects in the human point cloud model reconstruction are analyzed by experimental comparison.After an in-depth study,the Poisson surface reconstruction algorithm was finally adopted to construct the surface of the human point cloud model.4)Surface feature point calibration.After the 3D model of the human body is obtained by point cloud surface reconstruction,the height and 3D data of the 3D human body are obtained by measurement.The 81 feature points are calibrated on the surface of the standard human body model,and the measured height and circumference data are used to scale and transform the standard human body model to achieve rapid calibration of the 3D human body surface feature points.In summary,this paper obtained the 3D model of human body through 3D reconstruction,and used the 3D anthropometric method to obtain the subject’s body size data;combined with the obtained body size data and the standard human body model,the calibration of the human body surface feature points was realized,which helps to locate the contaminated parts of the body surface quickly.