Binocular Vision Sensor (Kinect)-Based Pedestrian Following Mobile Robot

Abstract

This paper present a pedestrian following mobile robot with binocular vision sensor. Because Kinect is one of the most inexpensive devices of depth-cameras, it is used in our application. Human skeleton is extracted by using Kinect, and the location of human is checked by projecting the three-dimensional (3D) pose of skeleton onto 2D screen. This 2D screen is separated into three parts, left, middle and right. Mobile robot rotates and translates according to the corresponding location of pedestrian. To make the robot move forward and backward, the distance between spine point and mobile robot is calculated. Finally, a real experimental result is used to validate our proposed method.

A Real-Time Tracking Algorithm for Human Following Mobile Robot using 3D Sensor

“…They are based on information about the shape or colour of objects. Commonly methods such as SURF, SIFT, HOG, Haar and LBP are applied in object recognition [6][7][8][9][10]. The recent development in machine learning and deep learning approach has a great impact on the accuracy of object detecting and tracking.…”

Untitled

“…They are based on information about the shape or colour of objects. Commonly methods such as SURF, SIFT, HOG, Haar and LBP are applied in object recognition [6][7][8][9][10]. The recent development in machine learning and deep learning approach has a great impact on the accuracy of object detecting and tracking.…”