Direct monocular vision algorithm based on deep constraints of point and line features fusion

文献类型： 外文期刊

第一作者： Bai, Keqiang

作者： Bai, Keqiang;Li, Xiuhong;Zhu, Yalan;Wang, Feiyan;Wang, Feiyan

作者机构：

期刊名称：OPTICS LETTERS （ 影响因子：3.3； 五年影响因子：3.2 ）

ISSN： 0146-9592

年卷期： 2025 年 50 卷 12 期

页码：

收录情况： SCI

摘要： The development of technology and the rapid increase of computing power have enabled the wide application of simultaneous localization and mapping (SLAM) in smart devices. Nevertheless, visual odometry based on the direct method exhibits inaccurate pose estimation in structured environments, because it ignores diverse line segment information, constraints of associated points, and estimated position information. Objective: This study aimed to address the issue of inaccurate pose estimation in structured environments for direct method-based visual odometry by proposing a direct monocular vision algorithm based on deep constraints of point and line features (DMVA-PLF), with the goal of improving pose estimation accuracy. Methods: The algorithm integrated line features from the environment into visual odometry through colinear and deep constraints, combined with historical pose information, to more effectively utilize image features and optimize the pose estimation process. Achievements: Experimental results demonstrated that the DMVA-PLF significantly improved pose estimation accuracy by efficiently leveraging image features, outperforming traditional methods in structured environments. (c) 2025 are reserved.

分类号：

  • 相关文献
作者其他论文 更多>>

微信小程序