Development of a new method for obtaining three dimensional body measurements from two dimensional

Since there is rapid increse in e-shopping, especially clothing, problems arouse and the most important one is unfit. In this research, to solve unfit problem, eliminate need for human body when obtaining body measurements, discard 3D body scanners, a novel method is proposed, to obtain 3D body measurements from 2D images. 2D images and their manual body measurements of 500 volunteers are processed, test group of 50 is separated at the beginning to be used at the end. Normal distribution of body measurement data is tested with five criteria of normality assumptions, boxplots, Kolmogorov-Smirnov and Shapiro-Wilk tests, proved to display normal distribution. Statistical selection of body measurements method is proposed to be shorter, precise, time-cost saving, correlation analyses are conducted. Multivariate multiple regression analyses are run with Model-1, Model-2 ve Model-3, Model-1 accuracy rate % 97,042 , artificial neural networks are run with Model-4 and Model-5, Model-4 accuracy rate % 97,097 .A new image processing software developed by us named ImPrNGF-1.0 is exercised, problems occured, two new developed softwares PyNGF-1 and ImPrNGF-2.0 are exercised, 240 threshold, sharp silhouettes are achieved. To obtain pixel coordinates from sharp silhouettes, two new softwares PixelList1.0 and PyNGF-2 are exercised, coordinate values are listed, then, subtracted from the one above and below in the list, 2D body measurements in pixels are obtained. Different mathematical models (circle, ellipse, and hippodrome formulae) are applied to convert to 3D body measurements, conversion coefficient C is developed to convert to centimeters since manual body measurements are in centimeters. It is concluded that ellipse circumference formula and 0.2596 cm/pixel C to be used. MMRA, ANN, and pixel studies are repeated with 50 test group, results are compared with manually taken body measurements, accuracy rate %98,490 . The proposed novel method to obtain 3D body measurements from 2D images is validated.