The Performance Analysis of Stereo Visual Odometry Assisted Low-Cost INS/GPS Integration System

• Main Authors: Li-HungChen, 陳麗虹
• Other Authors: Kai-Wei Chiang
• Format: Others
• Language: en_US
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/21585913885076633216

碩士 === 國立成功大學 === 測量及空間資訊學系 === 102 === In recent years, there are increasing applications of Inertial Navigation System (INS) with different aiding sensors such as Global Positioning System (GPS), odometer, heading rate sensor and so on. The INS/GPS integration system with odometer is the most commonly used in the land vehicular navigation area. In this research, a novel method is proposed for assisting low-cost INS/GPS integration system with Visual Odometry (VO) system to replace above sensors. The VO can estimate the movement of vehicle. When GPS signal vanishes, VO can be applied as the backup for GPS. The VO provides velocity information and updates the estimated position of the vehicle to reduce the INS accumulation errors then to improve the positioning accuracy. The tactical INS/GNSS (Global Navigation Satellite System) trajectory processed with RTS (Rauch-Tung-Striebel) smoother in the tightly coupled mode is used as the reference trajectory. The distances traveled (DT) by low-cost INS/GPS and VO modes are calculated. The trajectory comparison scenario is conducted in this research as well. In addition, this study investigates the accuracy variation in both monocular and stereo-vision VO system, respectively. The purpose is to verify the argument presented in related work. The settings of our experiments include different image sampling rates, different scenes and driving dynamics. According to the preliminary results, the position errors in three directions are reduced when GPS outage period is over two minutes and the improvement of 3D position RMSE is more than 40% by using the assisted information including VO position and velocity. Based on the preliminary results shown in the research, the stereo VO performs well and the performance of the proposed system is good in the GPS-hostile environment.