Construction Of Fabric Surface Model Based On Laser Scanning And Measurement Of Structural Parameters

Fabric surface analysis and measurement is very important for the textile manufacturing process and is traditionally based on human-labor which is time-consuming. What’s more, the efficiency and the accuracy of results depend largely on human. Image processing is a fundamental and conventional automatic technique to analyze fabric surface. However, image quality is influenced largely by ambient light, background light and optical properties of the surface. In this paper, a three dimensional modeling technique is used to reconstruct and analyze carbon-fiber fabric surface.Firstly, a dense points cloud is collected with the aid of 3D laser scanning system which is composed of a line structured-light sensor and a flexible coordinate measuring machine. In the scanning process, the quantity and the quality of points cloud data are rationally controlled by adjusting light condition, analyzing weave cycle diagram and selecting proper scanning scope.Secondly, original scanning points cloud is classified and preprocessed in different ways according to their distribution. Outliers, which are useless to describe features of fabric surface, are automatically deleted using the improved average Euclidean distance detection algorithm. In experiments, the 1.859% of average outliers are removed. On the basis of comparing classical isotropic and anisotropic filtering algorithms, inliers, which are within the allowable error and tend to usable points after necessary processing, are filtered using Laplacian smoothing, Taubin’s λ/μ algorithm and bilateral filtering algorithm, respectively. The experimental results show that λ/μ algorithm is appropriate to filter the carbon-fiber fabric surface mesh because it tends to preserve and enhance features while filtering.Thirdly, an ideal three dimensional model of fabric surface is reconstructed using Screened Poisson reconstruction algorithm. Scale-independent screened constrains including points constrains and gradient constrains are interpolated in the input points set. And screened Poisson equation is taken as the input equation of Marching Cubes (MC) algorithm. Cascadic multigrid method is used to solve screened Poisson equation. Asymptotic optimization method is given to relieve ambiguous problem in extracting isosurface. The results of comparative experiments demonstrate that Screened Poisson reconstruction algorithm is appropriate to reconstruct carbon-fiber fabric surface model. The experimental results show that the results are satisfactory according to both the effect of reconstruction and the running efficiency.Finally, the width of a yarn and fabric density has been automatically calculated using depth map and discrete wavelet techniques. The reconstructed model is projected onto a 2D plane using the parallel projection matrix. Then ideal depth-gray map of fabric surface is obtained with the aid of the gray transform method. In the depth-gray map, the gray distribution is decided by depth rather than by the light, the effects of ambient light, background light and other factors on the surface are removed to the greatest extent. Discrete wavelet technique is applied to separate the yarns from the depth map. The experimental results suggest that the reconstructed model is acceptable, and the measurement of structural parameters using depth map and discrete wavelet technique is reliable.