摘要
提出一种基于轮式里程计的全景惯性SLAM(simultaneous localization and mapping)初始化方法,用于地面测量系统、地面移动机器人中。在地面载体下,单目视觉惯性SLAM由于常加速或者纯旋转运动会出现退化问题,导致无法正常估计视觉尺度而造成系统初始化失败,而引入轮式里程计可以解决这一问题。首先,基于轮式里程计的双轮差速模型推导了两个视觉帧之间的轮式里程计旋转平移的预积分模型;其次,提出一个系统初始化方法,先分步初始化视觉尺度和IMU(inertial measurement unit)参数,再将全景相机、IMU、轮式里程计的观测值进行联合法平差得到精化后的初始化参数。实验结果表明,该算法能够得到较高精度的尺度信息,完成全景惯性SLAM系统的初始化流程。
This paper proposes an initialization method of panoramic-inertial SLAM(simultaneous localization and mapping(SLAM))based on wheel odometry,which can be used in MMS(mobile mapping systems)and ground mobile robots.Under the ground carrier,the monocular visual inertial SLAM will degenerate due to constant acceleration or pure rotation motion,which leads to the inability to estimate the visual scale and causes the system to fail to initialize and this problem can be solved by introducing wheel odometry.First,based on the differential model of the wheel odometry,the wheel pre-integration is derived which combines many measurements of wheels between two frames into a single relative motion constraint.Second,we propose a system initialization method,which first initializes the visual scale and IMU(inertial measurement unit)parameters step by step,and then integrates the observation values of the panoramic camera,IMU,and wheel odometry by perform combined adjustment to obtain refined initialization parameters.The experimental results show that the initialization algorithm can obtain high-precision scale information and complete the initialization process of the panoramic inertial SLAM system.
作者
姜帆
相恒茂
季顺平
JIANG Fan;XIANG Hengmao;JI S hunping(School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 430079,China;Shandong Provincial Institute of Land Surveying and Mapping,Jinan 250000,China)
出处
《测绘地理信息》
CSCD
2023年第4期21-24,共4页
Journal of Geomatics
基金
国家重点研发计划(2018YFB0505003)。
