Research On Clothing Measurement Based On Two-dimensional Images Based On Human Sample Databas

With the change of clothing market environment and the maturity of Internet technology,intellectualization and digitization have become a major trend in the development of clothing industry.The traditional clothing production and sales model is difficult to meet the fast-paced life needs of consumers.Therefore,clothing online shopping and personalized customization have become the main forms of people’s consumption.In 2020,the scale of China’s personalized customized clothing market will reach 200 billion yuan.The development of the market needs to be supported by technology,and anthropometric technology is the key to clothing online shopping and personalized customization.At present,anthropometry is divided into traditional manual measurement,non-contact three-dimensional anthropometry and twodimensional anthropometry.The traditional manual measurement method is inefficient and the accuracy is difficult to be guaranteed;The non-contact three-dimensional anthropometry has high accuracy,but the instrument is large and expensive,so it is difficult to popularize in practice.The two-dimensional human body measurement method can obtain the body size only by identifying the front and side images of the human body,which is convenient and fast,but the current measurement results do not meet the particularity of the individual body.Therefore,this paper makes an in-depth study on the two-dimensional image clothing measurement technology based on the self collected human body sample database.Firstly,human contour is extracted based on edge detection algorithm.By analyzing the function and detection effect of different edge detection algorithms,the adaptive threshold is used to replace the manually set double threshold in Canny edge operator,and then the improved Canny edge operator is combined with mathematical morphology to solve the problems of edge discontinuity and large noise in the image,and realize automatic human contour detection.Secondly,human feature points are detected based on joint point recognition algorithm and human proportion.This paper realizes the accurate recognition of human bone joint points based on neural network algorithm.For the unrecognized chest,waist and other parts,combined with the calculation of human body proportion,finally traverse the detection feature points in the contour image.This method solves the problems of false detection,missed detection and inaccurate location in the traditional feature point detection algorithm,and is suitable for groups with special upper and lower body proportion.Then,the length and girth parameters are calculated based on the human sample database.The length is calculated according to the distance between feature points.The girth adopts the regression analysis algorithm.Based on 122 groups of male data collected by ourselves,the fitting degree of the regression model is improved by about10% compared with that before classification by using the methods of correlation analysis,determination of independent variables and data classification.The average accuracy rate of each parameter within ± 1.5cm is 91%,which basically meets the requirements of clothing size.Finally,the design of two-dimensional image garment measurement software is realized.Open CV image processing and Tkinter graphical interface module based on python,as well as the front and side images of human body taken by mobile phone,complete the whole system process design of information management,image upload,image processing and result display.This paper studies the garment measurement technology based on the front and side images of human body taken by mobile phone.On the one hand,mobile phone measurement software can bring more convenient measurement methods to consumers;On the other hand,it can provide accurate size data for ordinary consumers and consumers with special figure,so as to meet the needs of clothing online shopping and personalized customization.