一种基于惯性传感器Roll角和Pitch角的单目视觉同步定位与地图.docVIP

一种基于部分惯性传感器信息的单目视觉同步定位与地图创建方法( 顾照鹏，董秋雷 (中国科学院自动化研究所模式识别国家重点实验室 北京 100190) (zpgu@nlpr.ia.ac.cn) 摘 要: 传统的单目视觉同步定位与地图创建（MonoSLAM）方法很难处理累积误差问题，如何有效利用惯性传感器输出的运动信息辅助SLAM系统抑制累积误差是MonoSLAM研究中的一项重要内容. 在惯性传感器输出的三轴方向角中横滚角(Roll)和俯仰角(Pitch)的精度较高，而偏航角(Yaw)的精度相对较低，我们在实验中发现，如果在SLAM系统中直接使用惯性传感器输出的偏航角信息，不但无法有效抑制SLAM系统的累积误差，反而会进一步增大系统误差、降低SLAM系统的稳定性.针对这种情况，本文提出了一种基于惯性传感器横滚角和俯仰角角的MonoSLAM方法. 本文首先探索了基于横滚角和俯仰角的系统标定问题，即如何仅仅利用惯性传感器输出的横滚角和俯仰角信息标定惯性传感器与摄像机组成的定位定姿系统. 然后将惯性传感器自身的偏航角作为系统状态向量的一个分量，利用扩展卡尔曼滤波器实时地估计状态向量，进而实现实时鲁棒地同步定位和地图创建.本文方法可以有效地抑制SLAM系统运行过程中产生的累积误差，并降低惯性传感器测量误差对SLAM系统稳定性的影响. 大量对比实验验证了本文方法的有效性和实时性. 关键词: 同步定位与地图创建;惯性传感器;扩展卡尔曼滤波器 中图法分类号: TP391 Monocular SLAM Based on Partial IMU Information Gu Zhaopeng and Dong Qiulei (National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, Beijing 100190) Abstract: It is difficult for traditional MonoSLAM (Monocular Simultaneous Localization and Mapping) methods to deal with the accumulated errors effectively, so how to fuse data from an IMU (Inertial Measurement Unit) with data from a MonoSLAM system to reduce such errors becomes an important issue currently. We observed that, among the three Euler angles from an IMU, Roll and Pitch are more accurate than Yaw. Then if Yaw is used directly in a MonoSLAM system, the accumulated errors will increase but rather than decrease. To address this problem, a new monocular SLAM method based only on Roll and Pitch output from an IMU is presented. First, a calibration algorithm for the IMU-vision system using only Roll and Pitch is proposed. Then, the EKF (Extended Kalman Filter) is adopted to estimate the state vector comprising the camera position, Yaw of the IMU and the locations of features, and the SLAM system based on only Roll and Pitch is implemented robustly in real time. Experimental results show that the proposed method can reduce the accumulated errors effectively and alleviate the negative influence of the IMU’s measurement errors on the system’s st