Gait analysis with Drone

Abstract

This project presents a flexible and easy way to perform analysis on some most important parameters (i.e. step length, stride length, step frequency, speed) of the gait by drone in an outdoor environment. I developed an iOS application by DJI SDK to connect to the drone through the remote controller. The drone films a video of the target walking, at the same time recording the flying height and pitch angle of camera. From the video, the points of human body can be extracted by OpenPose library which is a deep learning model to estimate the key points of joints in human body. The step frequency is estimated from those key points. Simply by triangulation, the actual length is calculated from the frame of video to estimate the stride length. Unlike traditional way of measurement, this technique requires only the height of camera and pitch angle. Then the running speed can be calculated from the step frequency and step length.