Optimal Resolution Of3D Garment Model Based On Geometric Deviation And Curvature Analysis

3D garment models of scanner or garment CAD software had gained more and moreapplications such as online fitting and fit evaluation. Since the data volumes of garment modelsoften beyond that required for the real time view, garment models were simplified to meet therequirement. However, simplification of garment models inevitably resulted in loss of visual effect.Current researches are mostly focused on how to improve the efficiency and accuracy ofsimplification algorithms, but few studied the relationship between ratio of simplification andvisual effect of garment models, but the visual effect of garment models is the most concernedattribute for users. Hence, it is necessary to study the relationship between the data volume of3Dmodel and its visual effect, and find the optimal resolutions with least amount of data that canshow details of garments.Firstly, some daily wear clothing were selected as samples, and then scanned with a Kinectscanner based on Kinect Fusion technology. In the scanning process, a circular guide bracket assupport was adopted to obtain original3D model of garments. Secondly, a visual effect evaluationsystem which composed of geometric deviation and curvature distribution was established toevaluate the simplified garment models, analyze the relationship between ratio of simplificationand visual effect, and get the optimal resolutions of garment models. Herein, mesh models andpoint-cloud models were took as samples, corresponding simplification algorithm for quadric errormetric (QEM) and resampling method of Fourier theory, respectively. Finally, a subject evaluationexperiment was designed to verify whether the optimal resolutions of garment models areacceptable by human eyes or not.After the above research process, results were got as follows. Based on visual effectevaluation, to most mesh models or point-cloud models, a90%simplification would notsignificantly affects visual effect; Compared to the original models, the models with optimalresolution can reduce CPU usage; The relationship between triangle number and area of garmentmodels obeys empirical formula y=1981x+20090, where x represents area of original3D garment models and y represent triangle number of3D garment models. The optimal point-clouddensity of point-cloud models is about8195points/m2; When the ratio of simplification is lessthan90%, visual effect of garment models did not changed significantly, whereas the visual effectwould decreased dramatically when the ratio is greater than90%; As the rate of simplificationincrease, the variation of geometric deviation and curvature distribution was almost same in meshmodels, while was much different in point cloud models.Above results provided a theoretical basis for3D garment models in real time view. It means90%data volume of garment models can be simplified, which greatly accelerate the speed of realtime view of garment models. At the same time, the empirical formula of optimal number oftriangles and optimal point-cloud density provide support to the scanning of garment. It means thenumber of triangles and points can be calculated according to the areas of garment models,thereby regulating scanning accuracy and improving the utilization of scanning resources, andfinally reducing the scanning time and improving scanning efficiency.