Research On SLAM Algorithm Of Indoor Mobile Robot Based On Monocular Vision And Odometer

Simultaneous localization and mapping(SLAM)provides a better solution for indoor mobile robots,and is the core of autonomous navigation for indoor mobile robots.With the development of 3D vision,SLAM algorithm based on monocular vision has become a research hotspot.However,monocular slam depends on the scene features,and the lack of texture and fast motion may lead to tracking failure.Moreover,there are scale initialization problems and scale drift when using monocular vision sensor.The output frequency of odometer is high,and odometer can estimate the position and pose with scale in real time.The short-term estimation is accurate,but the cumulative error of long-term estimation is large,and the long-term estimation of monocular vision sensor is more accurate.Because of the obvious complementarity of the two sensors,monocular vision and odometer are fused to solve the localization problem of indoor mobile robot.The specific research contents of this paper are as follows:(1)The visual slam system is deeply studied,and its development and research status are briefly described.At the same time,the pose transformation theory of rigid body motion,camera calibration technology and odometer motion model are studied,which provides a theoretical basis for the construction of slam system based monocular vision and odometer.(2)Aiming at the problem of external parameter calibration of monocular vision sensor and odometer,an effective algorithm using only checkerboard is proposed.The algorithm uses checkerboard to solve the pose of camera coordinate system relative to the world coordinate system,and then uses the least square method to fit the x-axis direction vector of odometer coordinate system.Since the y-axis vector is known,the z-axis direction vector is obtained in the right-handed coordinate system.The rotation matrix composed of these three direction vectors is taken as the pose of odometer coordinate system relative to the world coordinate system,and the pose of camera coordinate system relative to odometer coordinate system is obtained by matrix product.(3)According to the characteristics of monocular vision sensor and odometer,a SLAM algorithm based on monocular vision and odometer is proposed.The algorithm includes system initialization,scale recovery,front-end odometer and back-end optimization.Monocular vision has the problem of scale initialization.In this paper,odometer is used to restore the visual scale in the initialization stage.In the indoor scene,the pose of the mobile robot is constrained,which includes the feature point re projection error constraint,odometer constraint and prior plane constraint,which makes the pose estimation more accurate.Finally,experiments and precision analysis are carried out to verify the effectiveness of the proposed fusion algorithm.