摘要
针对低纹理环境下或相机快速旋转时单目相机定位中存在的精度和鲁棒性差的问题,给出了一种集成线特征的单目视觉惯性里程计算法(Visual Inertial Odometry,VIO),并采用Dog-Leg方法代替Levenberg-Marquard方法进行非线性求解,降低了运算量。基于实测数据的实验结果也表明,相比于VINS-Mono算法和OKVIS-Mono算法,所给出的定位方法可以获得更高的定位精度。
In this paper,the accuracy and robustness of monocular camera positioning in low-texture environment or camera rotation are presented.A monocular visual inertial odometry(VIO)method with integrated line features is presented.The Dog-Leg method is used instead of the Levenberg-Marquard method for nonlinear solution,which reduces the amount of computation.The experimental results based on the measured data also show that the positioning method can achieve higher positioning accuracy than the VINS-Mono algorithm and the OKVIS-Mono algorithm.
作者
周扬
王勇
徐国亮
ZHOU Yang;WANG Yong;XU Guo-liang(Jiangsu Automation Research Institute of CSIC,Lianyungang 222061,China)
出处
《指挥控制与仿真》
2020年第1期47-53,共7页
Command Control & Simulation
