3D Size-Estimation based on the Geodesic Distance Measured by Photogrammetric Scanning Device

Abstract

This paper introduces an innovative approach for estimating important body measurements according to the depth information extracted by photogrammetric scanning devices, such as Microsoft Kinect II, and verifies its applicability and efficiency through applying it to an experimental use-case concerning classification based on size, to be implemented in virtual fitting rooms. The mathematical framework consists, mainly, in constructing a projection from the screen coordinates associated with the pixels onto the real-world ones, and then integrating the geodesic distance gradients throughout a path connecting the beginning and end points. Besides, in order to reduce the inaccuracy through excluding the possible high-frequency noise, and to obtain more realistic measurements through minimizing the discrepancy between the original path and the one taken into account in the integration module, the path is fitted to a smooth curve in a separate stage before performing the geodesic-distancecalculation iterations. In the end, in order to verify the accuracy and preciseness of the proposed technique, it is applied for obtaining numerous body sizes, where for ensuring that the correct measurements are achieved, imaginary landmarks are placed on the bodies of the subjects. The results show that the accuracy is less than half a centimeter using Microsoft Kinect II.