DEVELOPMENT OF A METHOD FOR MEASURING A PEDESTRIAN TRAJECTORY USING SLAM BASED ON FIRST-PERSON VIEW IMAGES

We developed a method for measuring a pedestrian trajectory using a depth camera, which provides both infrared and visible light images. From these first-person view images, a SLAM technique generates the three-dimensional map around the camera and estimates the position of the camera using the map. We apply this method to measure the trajectory of a pedestrian avoiding an object, and the result gives a reliable walk trajectory and a detailed map including the object shape. We also measure two-pedestrian passing each other and show that this method can simultaneously measure the trajectories of multiple pedestrians.