摘要
Abstract
Visual inertial odometry based on point features performs poorly in unstructured and other complex scenes,while line characterization have rich geometric information in the environment compared to point features.The article aimed to utilize line characterization to improve the positioning accuracy of the system.However,the extraction and tracking process of line characterization is very time-consuming,and the wrong detection and matching of line characterization affects the accuracy and robustness of the system.In this regard,the article proposed an improved EDLines line segment extraction algorithm,which improved the quality of the line feature extraction by short line rejection with line segment complementation and uniform distribution of line features.The experimental results showed that the algorithm improved the efficiency and accuracy of visual inertial odometer localization.关键词
视觉惯性里程计/线特征/制图技术/点特征/定位精度Key words
visual inertial odometry/line characterization/mapping technique/point features/localization accuracy分类
信息技术与安全科学
