３次元点群データを用いた歩行時の全身計測システム

抄録

Along with the aging of the population in recent years, fall accidents due to declining walking ability of the elderly are increasing. In clinical practice such as medical institutions, people with knowledge of medicine such as physiotherapists and doctors often qualitatively evaluate and judge the individual walking abilities from their experiences and tendencies. However, when the number of subjects is large, or when quantitative evaluations such as joint angle and center of gravity are performed, it is difficult in the current system. In this research, we propose a quantitative measurement system using a skeleton model of a depth sensor (Kinect v2). Kinect can estimate the position of a person's joint using depth information. However, due to the measurement range of the skeleton model and constraints on inaccuracy, correction for 3-D analysis is necessary. In this report, we aim to track data of turning without missing while using two Kinects, while interpolating each other points. We use iterative closest point (ICP) algorithm to calibrate the axis of coordinates of two Kinects. We created an algorithm that switches estimation methods according to observation values. In addition, comparison of precision was made by comparing with Vicon which was measured at the same time.