.png){kind=icon}
Return
Multi-source constrained ORB-SLAM3 algorithm
.png){kind=icon}
PRE
AI
DOI：10.1088/2631-8695/ae602a.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
To address the trajectory drift and low localization accuracy of traditional simultaneous localization and mapping in dynamic environments, along with the loss of effective static features caused by over-pruning in existing dynamic methods, this paper proposes a multi-source constrained ORB-SLAM3 algorithm. The core innovation of this system is a refined multi-source constraint fusion strategy. This strategy selectively retains high-quality static feature points within dynamic regions by jointly modeling semantic priors extracted via YOLOv11n, spatial affiliations, and kinematic residuals while employing an optimized weight allocation mechanism. We also design a decoupled anisotropic process noise covariance matrix for the Kalman filter to suppress bounding box tracking jitter and apply a bidirectional optical flow and Sampson validation algorithm to rigorously eliminate residual outliers in static areas. Experiments on the TUM RGB-D dataset demonstrate that this approach reduces the root mean square error of the absolute trajectory error by up to 97.14% in highly dynamic sequences compared to the original ORB-SLAM3 algorithm, achieving robust and high-precision pose estimation under dynamic interference.
Keywords:
ORB-SLAM3
YOLOv11n
Kalman filter
optical flow-Simpsons algorithm
multi-source constraint fusion algorithm
Journal
E
IF:
1.6
Papers:
2.1K
Citations:
0
Organization
Cited Papers
No cited papers available