Characterization of clothing ease allowance based on the 3D reconstruction and point cloud registration of multi-view images
To analyze the amount of clothing ease allowance under different postures of the human body,the three-dimensional(3D)point cloud of the nude body and clothing was obtained by 3D reconstruction based on multi-view images.Using the human pose estimation model of OpenPose,we estimated the joints and reversely projected them to the 3D space to get the coordinates of the matching points for the coarse registration of the two layers of point clouds of the nude body and clothing.We realized the accurate registration of the two layers of point clouds based on the uncovered human body point clouds by the iterative closest point(ICP)algorithm,and verified the registration model,building a 3D dressing model.Based on the model,we computed the nearest-neighbor distance between the two layers of point clouds,and defined the distance as the ease allowance.The results showed that after two-step registration,the average distance and root mean square error of the two layers of point clouds in the uncovered area of the torso were 2.859 mm and 3.260 mm,respectively,and the maximum average distance and root mean square error of the two layers of point clouds in the arm area were 4.018 mm and 4.735 mm,respectively,indicating good accuracy in registration results.This method can accurately obtain the size and distribution of clothing ease allowance for different postures and provide key technical means for analyzing the relationship between clothing ease allowance and movement,evaluating comfort,establishing heat transfer models,and evaluating virtual fitting effects.
NETL
NSTL
万方数据
