Multi-view Anthropometric Measurement For Garment Design

Human body size has a wide range of applications in the fields of virtual fitting systems,ergonomics,and clothing science.It is also the basic technical basis for measuring a country's production capacity.As clothing is the most intimate contact with the human body,body size data is very important.Aiming at the problems of expensive non-contact measurement equipment and the large space occupied by the equipment,exploring non-contact measurement systems suitable for the Chinese clothing industry is of high practical value.Currently,human body measurement methods include contact and non-contact measurement.Non-contact measurement is divided into active and passive.In view of the shortcomings of the active three-dimensional scanning measurement method and the passive single-viewpoint and dual-viewpoint measurement methods,we have designed a multi-view body measurement system.The main research contents of this article include the following three parts:1.The multi-view anthropometric system designed in this paper is used to collect the subject's image sequence at the same time,and the trained GSS-PSPNet neural network is used to perform semantic segmentation on the body measurement part to obtain an image containing only the measurement part.2.The image containing the measurement part is subjected to image matching operation,combined with the specific structural characteristics of the identified points,and matched using a color space clustering algorithm to obtain the three-dimensional coordinate value of the characteristic point of the measurement part,that is,the position in the European space.3.Using the principle of European-style spatial coordinate conversion,the points in the coordinate system established by the three groups of binocular are unified in the same coordinate system.Finally,the marked points of the measurement site are fitted to obtain a fitting curve.The length value is the size of the measurement site.Our system overcomes the problems of the complexity of traditional manual measurement methods,the heavy workload,the psychological pressure on users caused by the active light source emitted by the active measurement method,the expensive equipment,and the low accuracy of single view and double view measurement systems.We tested the subjects of different genders and shapes respectively.The absolute error range of measurement results was [0.78,1.58] cm,and the maximum relative error rate was 1.77%.The measurement error of waist and bust of human body met the China national standards GB/T 2664-2017,GB/T 2665-2017,and textile industry standard FZ/T 81006-2017,FZ/T 81004-2012.The experimental results show that our system has good accuracy and reliability.