Stereo Visual SLAM Using Point And Line Features

Simultaneous Localization and Mapping(SLAM)is one of the key technology in the field of robotics,which has a wide range of applications.Due to the unique advantages of visual sensors,visual SLAM has become a research hotspot in recent years.Point feature is the most widely used feature in visual SLAM,and it is more mature in feature extraction,matching and representation.However,point features are more dependent on the environment,and are not performing well in textureless scenes and etc.In man-made structured environment,there are plenty of line features.In fact,line features and point features provide complementary information about the image.Meanwhile,the line feature is a higher level primitive,and the environment map constructed by line segments has more intuitive geometric information.In this thesis,we propose a visual SLAM method based on point and line features extracted from the stereo camera,which can be used to locate and estimate the robot in indoor and outdoor environments.The main research work includes:1.An improved extraction and matching method for line features is introduced to robustify data association by merging the broken line segments,and using the constant velocity model and Liang-Barsky line segment clipping algorithm to predict the camera observations.2.In the back-end of the proposed visual SLAM,we construct the graph model using point and line features.And we employ the orthonormal(minimal)representation to parameterize lines and analytically compute the corresponding Jacobians.3.We design and implement a complete visual SLAM system using both point and line features,which includes stereo matching,frame tracking,local mapping,bundle adjustment of both line feature and point feature,as well as point-line based loop detection.Extensive experimental results are presented to validate its performance.