Research On Mobile Robot Visual Odometer Based On Point Line Feature Fusion

With the wide application of mobile robots,its precise positioning technology has gradually become a research hotspot as one of the core technologies in this field.Depending on the positioning of a single sensor,it can not meet the precise positioning requirements in complex environments.Therefore,the multi-sensor information fusion positioning method has become the main method for pose estimation of mobile robots today.At present,more and more applications are based on the fusion of visual information and inertial navigation information to achieve accurate position and pose estimation.In the research of visual-inertial navigation odometer,image geometric information is usually obtained by feature point matching,but the effective feature points extracted in low-texture scenes are less,which may lead to inaccurate positioning.In the artificial environment,there are a large number of line features,which can make up for the lack of point feature information.Therefore,it is of great practical value to combine the point line features to improve the positioning accuracy based on the fusion of visual inertial navigation information.The method studied in this paper is to fuse the visual and inertial navigation information in a tightly coupled filtering manner,and estimate the geometric information of the image by combining the point-line features,so as to achieve accurate pose estimation under the condition of information fusion.This paper mainly studies from the following aspects.Firstly,the composition and related principles of the inertial measurement unit IMU and several methods of pose description are described.The projection model of pinhole camera and the relationship among four coordinate systems used in camera calibration are introduced.The camera calibration experiment and the camera-inertial calibration experiment are carried out.The internal and external parameters of the camera and the relative pose between the camera and IMU are obtained,which laid a foundation for the follow-up research.Then,the extraction,description and parametric representation of the line features are introduced in detail,and the observation model of the spatial line is constructed to correct the prior information of the inertial navigation.After that,the visual inertial navigation localization algorithm based on point-line feature is deduced in detail.Finally,the EOURC data set is used for experiments to estimate the camera's motion position and pose,and the comparison map of the estimated trajectory and the real trajectory is outputs.The error analysis shows that the visual inertial navigation information fusion localization algorithm based on point-line features effectively improves the positioning accuracy.