Robot localization via a RGB-D camera

• Main Author: Yan, Hui
• Other Authors: Hu Guoqiang
• Format: Theses and Dissertations
• Language: English
• Published: 2016
• Subjects: DRNTU::Engineering::Electrical and electronic engineering
• Online Access: http://hdl.handle.net/10356/65903
• Institution: Nanyang Technological University

Robots play an important role in industry, military applications and even our daily life. One of the greatest advantages of robots, comparing with computers, is the mobility, which brings more application areas to robots. The localization problem is a fundamental and classic basis for most vehicles and robots, including two topics: localizing the objects in the environment and mobile robot self-localization. In this dissertation, self-localization of mobile robots is considered. Instead of using traditional global positioning system (GPS) or inertial measurement units (IMU) techniques, localization algorithms are utilized to find the position of the mobile robots. Given recent advances in machine vision and image processing technology, computational technology, and control theory, one of the approaches to solve the self-localization of indoor mobile robots is to utilize a vision system. Generally, these techniques localize robot by matching salient features in the space detected by vision systems. This project uses a RGB-D camera as the vision sensor. It aims to develop a vision-based solution using a RGB-D camera for localization of a mobile robot. This dissertation introduces the characteristics and work scenarios of Kinect, giving a basic knowledge of RGB-D camera for reference. Methods of feature tracking and RGB-D camera calibration are also stated. In addition, simulation of two specific localization algorithms are carried out and reasons causing errors are analyzed. Finally, the effect of depth information of RGB-D camera is discussed