Research on the Effectiveness of Monocular Visual SLAM Depth Estimation Base on Improved ORB Algorithm
-
- Wang Jiwu
- School of Mechanical and Electronic Engineering, Beijing Jiaotong University
-
- wan Weipeng
- School of Mechanical and Electronic Engineering, Beijing Jiaotong University
説明
The application of monocular vision to measure the depth information of image feature points is one of the important tasks of monocular vision slam. Triangulation is a method often used to measure the depth information of feature points, but in the actual application process, the uncertainty of feature point matching will cause greater depth uncertainty. This paper proposes an improved ORB feature point extraction strategy, combined with the quad-tree model to achieve the homogenization of feature points. The feature points matching method uses the brute force matching method, RANSAC filters the matching point pairs, and obtains better matching results for depth estimation. Experiments show that the improved feature point extraction and matching method can effectively obtain camera pose estimation value and depth estimation value. And in this way, the accuracy of the estimated value is improved .
収録刊行物
-
- 人工生命とロボットに関する国際会議予稿集
-
人工生命とロボットに関する国際会議予稿集 27 318-322, 2022-01-20
株式会社ALife Robotics
- Tweet
詳細情報 詳細情報について
-
- CRID
- 1390854717506827904
-
- ISSN
- 21887829
-
- 本文言語コード
- en
-
- データソース種別
-
- JaLC
- Crossref
-
- 抄録ライセンスフラグ
- 使用不可
