An integrated 3D point cloud and 2D image-based approach for generating individualized female prototype pattern

Abstract

To promote online garment customization and virtual fitting, how to obtain sufficient body shape data easily and quickly to generate personalized tailoring patterns is still a challenge. In this study, an integrated 3D point-cloud and 2D image-based approach for generating individualized female prototype patterns was proposed, and an automatic pattern generation system based on body height, weight, and images was developed. A feature point recognition method was used to measure 14 body parameters, including the width, thickness, and angles at characteristic body landmarks. Multiple linear regressions between these parameters and the coordinates of 20 pattern measurement points were then performed. For the extracted 2D body parameters, the error for more than 80 % of the samples were within 1 cm by comparing with the 3D measurement data. Moreover, the predicted accuracy of the generated individual patterns reaches 91 % overall, demonstrating the reliability of this system. Therefore, this method can improve the efficacy of automatic garment production, and meet the growing requirements of sustainable development and individualized consumption.