A novel video-based method using projected light to measure trunk volumes during respiration

Abstract

This paper proposes and evaluates an innovative video-based method for measuring the trunk volume during respiration, using projected light and surface reconstruction. The method consists of the following main steps: (a) to project a grid of circular light markers on the anterior and posterior human body trunk surface during breathing, (b) to register the subject's trunk surface using two pairs of pre-calibrated digital video cameras, (c) to segment the video stream and track the projected markers using pre-processing techniques, morphological operators and detection algorithms, (d) to label the corresponding markers in the video sequences registered by each pair of stereo cameras, (e) to reconstruct the 3-D coordinates of all markers, (f) to reconstruct the surfaces from the unordered cloud of points using the Power Crust method and (g) to calculate the trunk volume in function of time using the divergence theorem. The evaluation of the method was based on two experiments. (1) Comparison of the volume of a trunk model (mannequin) by immersion and using the proposed optical method. (2) Analysis of the applicability of the method for measuring a subject’s trunk volume during a vital capacity respiratory manoeuvre. The results showed that the method was able to automatically measure more than 2000 projected points per image and to provide a very detailed representation of the subject's trunk. The relative accuracy of the volume measurement was estimated to be better than 3%. The analysis of the experiments revealed that signals coherent with the respiratory cycles could be identified through this method. In conclusion, the method based on light projection and surface reconstruction provides an accurate, non-invasive and useful means to calculate human trunk volumes during breathing.

Keywords:

• biomechanics
• Power Crust
• videogrammetry
• volume
• surface reconstruction

Acknowledgements

This research was supported by FAPESP (00/01293-I), CNPq (451878/2005-1; 309245/2006-0; 473729/2008-3; 304975/2009-5) and CAPES.