惯性/双目视觉里程计深组合导航方法  

INS/Stereo Visual Odometry Deeply Integrated Navigation Method

作　　者：逯建军[1] 任晓军 孙伟 郭元江 李群 

机构地区：[1]海军装备部,北京100000 [2]海军驻航天某院军代表室,北京100000 [3]北京自动化控制设备研究所,北京100074

出　　处：《导航定位与授时》2016年第3期37-43,共7页Navigation Positioning and Timing

基　　金：总装预研课题(51309030203)

摘　　要：针对现有视觉里程计测量噪声大、匹配精度低、实时性差的问题,研究一种基于Kalman滤波器的惯性/双目视觉里程计组合导航方法。在视觉里程计中引入惯性导航信息,辅助完成实时图截取、搜索区预测、输出速度校正等功能,提高视觉里程计的测量精度与计算速度。利用Kalman滤波器,实现视觉里程计对惯导累计误差的修正,提升组合导航系统的导航精度。车载试验结果表明,惯性/双目视觉里程计深组合导航的实时定位精度优于0.5%D(CEP),具备工程应用条件。To resolve the problem of large measurement error,low precision of matching,poor real timing,the INS / stereo visual odometry integrated navigation method based on Kalman filter was designed. The INS is used to help the stereo visual odometry faster and more accuracy by the real-time images capture,search area prediction and measuring velocity adjusting. By the use of kalman filter,visual odometry can correct the accumulated error of INS to improve the precision of integrated navigation system. The vehicle-carried result shows that this method improves the precision of INS / stereo visual odometry deeply integrated navigation to 0. 5%D（ CEP） and possess the condition of engineering use.

关 键 词：双目视觉里程计 惯性导航 组合导航 

分 类 号：U666.1[交通运输工程—船舶及航道工程]