Anthropometric Measurements with 2D Images


Erturk R. A., Karnasak M. E.

7th International Conference on Computer Science and Engineering, UBMK 2022, Diyarbakır, Turkey, 14 - 16 September 2022, pp.251-256 identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/ubmk55850.2022.9919504
  • City: Diyarbakır
  • Country: Turkey
  • Page Numbers: pp.251-256
  • Keywords: anthropometry, body segmentation, pose estimation
  • Istanbul Technical University Affiliated: Yes

Abstract

© 2022 IEEE.In this work, we designed and provided a proof-of-concept study for a novel system that takes several anthro-pometric measurements (bust, waist, and hip circumferences) simultaneously using only two (frontal and side) 2D images of a human subject. The system has two components: a specific camera setup with lasers and image analysis software. Towards this purpose, we compare body measurements of the proposed system and manual measurements on a limited number of subjects. For automatic measurements, we took one frontal and one side image from each subject. Body segmentation and pose estimation are applied to these images using pre-trained deep neural networks. Using the laser positions on the image, pixel sizes are estimated in terms of physical length (ie. centimeter). Using physical widths of the bust, waist, and hip in the images, their circumferences are estimated automatically. On three subjects, we obtained less than 10% measurement error. We concluded that anthropometric measurements could be obtained using a camera and laser setup. However, the number of subjects should be increased with a more precise laser to determine a better margin for measurement error.