| In the background of the prosperity of e-commerce,people buy clothing online more and more frequently,but there are more and more returns and exchanges due to inappropriate clothing sizes and the bad shopping experience increases.Therefore,the use of simple equipment measurement quickly and conveniently collects and calculates the accurate data of key parts of the body in real time to quickly find the appropriate clothing size for online clothing purchase and offline mass human body clothing customization,which is in great demand in the current market.Based on the characteristics of portable image acquisition equipment,this paper selects SLR cameras and mobile phone cameras with high retention rate and relatively low cost in daily life for human body image collection.This paper designs and implements a real-time non-contact automatic measurement system of human body size based on 2D images.By collecting the unmarked frontal image,side image and background image of the test subject,real-time automatic calculation of human body size and recommendation of clothing size model are recommended to meet the needs of scenes of online purchase of clothing and offline mass measurement.The concluded work is summarized as follows:(1)Calibrating the camera by Zhang Zhengyou’s plane calibration method and calculating the distortion parameters and correct the image to verify the feasibility of directly applying the collected unmarked human images to image processing and size calculation based on the comparison of pixel coordinate values and actual measurement values.It simplifies the process of automated calculation and reduces the background requirements and application threshold of this measurement technology.(2)By comparing the actual processing effects of the five filtering algorithms of box filtering,mean filtering,Gaussian filtering,median filtering,bilateral filtering and Sobel operator,Canny operator,Laplacian operator,the appropriate algorithm is selected.Constructed complete processing steps such as image graying,image difference,image binarization,image optimization based on morphological operations,contour detection,automatic size extraction and calculation.(3)Using the Haar detector for face detection,eye detection,and pupil center extraction to determine the data of pixels that the pupillary distance occupies and determine the pixel equivalent of the image based on the actual pupillary distance value of the input subject.The method of automatically extracting and calculating the height,shoulder width and waist circumference of the human body contour features.The accuracy of the results of this paper was verified by the comparison test of human body characteristic data of different shapes and the 3D point cloud data collected by Structure Sensor 3D scanner.(4)Using the C# programming language to develop a real-time automatic calculation software for human body size based on 2D images.According to the calculated gender,height,shoulder width and waist circumference of the tested person,the recommended reference size model is output according to the size chart. |
