一种基于YOLO的面向室内动态场景的VSLAM方法
By combining YOLO object detection and multi-view geometry to filter dynamic feature points, the accuracy problem of visual SLAM algorithm in dynamic environments was solved, achieving an accuracy improvement of 96.58%.
CN116758116BActive Publication Date: 2026-07-17NANCHANG UNIV
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NANCHANG UNIV
- Filing Date
- 2023-06-09
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Existing visual SLAM algorithms are not accurate in dynamic environments. Geometric methods are prone to over- or under-identification, while deep learning-based methods are time-consuming or have excessively high algorithm complexity.
Method used
The YOLO object detection algorithm is used to obtain prior semantic information of objects. Dynamic feature points are then filtered and eliminated by combining depth thresholding and multi-view geometric methods to improve accuracy.
Benefits of technology
The accuracy of the SLAM algorithm in indoor dynamic scenes has been improved, achieving an accuracy improvement of 96.58%.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN116758116B_ABST
Abstract
本发明提出了一种基于YOLO的面向室内动态场景的VSLAM方法，包括1）采集彩色图像和深度图像，输入SLAM系统；提取特征点及深度，并将特征点ID与深度值一一对应；2）通过YOLO目标检测算法检测图像中的物体，并获取其检测框；3）对于动态物体，将检测框中的特征点进行基于深度的聚类，将动态特征点进一步分离出来；4）对于潜在的动态物体，利用多视图几何原理结合深度阈值判断其动态性，当该物体被判断为动态物体时，认为其检测框内的特征点均为动态特征点；5）将所有动态特征点剔除，其余特征点输入SLAM系统，进行后续跟踪、建图和回环检测线程。与传统的视觉SLAM算法对比，本发明方法的精度可提升至96.58%。
Need to check novelty before this filing date? Find Prior Art