Using a single RGB camera to obtain accurate body dimensions rather than measuring these manually or via sophisticated multi-camera or laser-based sensors, has a high application potential for the apparel (fashion) industry. We present a system that estimates upper human body measurements using a set of computer vision and machine learning technologies. In a nutshell, the main steps involve: (1) using a portable camera (such as with a smartphone); (2) improving the image quality; (3) performing a calibration step; (4) extracting features of the body from the image; (5) indicating markers on the image semi-automatically; (6) producing refined final results.We experimented with the system on a sample of participants. The results for the upper human body measurements in comparison to the main manual method of tape measurements show ±1cm average differences, which is a good enough result for a number of applications.

In the present paper, ellipse-like approximations are considered with the aim of minimizing the difference between the results of direct and software measurements. The human body can only be approximately represented by elliptic cross sections, and these can vary for each individual. We show that better results are obtained by adapting the fit to the body shape based on a simple criterion. Based on this study, we have selected two mathematical models to estimate upper human body circumferences.The result is a fully trained system that can choose the best ellipse equation according to the human body shape to calculate human body circumferences.