A Garment Pattern Generation Framework Based on the Surface Flattening of a 3D Point Cloud

To achieve personalized garment production based on a three-dimensional (3D) point cloud, a garment pattern generation framework using surface flattening is proposed. 3D point cloud data of 190 subjects were preprocessed to achieve 17 dimensions by automatically locating five landmarks, including side neck point, front neck point, shoulder point, bust point, and waist point, by using the maximum distance method. 2D initial patterns were generated from processed point cloud data through structural line design and surface flattening. In total, 145 standardized garment patterns were justified according to the correction rules of a Japanese Bunka-Style Bodice Block to conform to standard structures. A 2D coordinate system containing 23 feature points, three auxiliary lines, and seven auxiliary points was established for pattern measurement. Finally, personalized pattern generation rules were determined by backward stepwise regression analysis between body measurements and pattern parameters. The results show that the mean error of five landmarks is below 2.49 mm, validating the accuracy of 3D point cloud-based automatic landmark localization, and the generated pattern errors of five key parameters meet standard tolerances. Meanwhile, virtual fitting effects of the generated patterns are satisfactory with low and evenly distributed pressure. The proposed garment pattern generation method meets personalized garment fit requirements and provides a feasible technical path for garment production in the fashion industry.