An Adaptive ORB-SLAM3 System for Outdoor Dynamic Environments

Abstract

Recent developments in robotics have heightened the need for visual SLAM. Dynamic objects are a major problem in visual SLAM which reduces the accuracy of localization due to the wrong epipolar geometry. This study set out to find a new method to address the low accuracy of visual SLAM in outdoor dynamic environments. We propose an adaptive feature point selection system for outdoor dynamic environments. Initially, we utilize YOLOv5s with the attention mechanism to obtain a priori dynamic objects in the scene. Then, feature points are selected using an adaptive feature point selector based on the number of a priori dynamic objects and the percentage of a priori dynamic objects occupied in the frame. Finally, dynamic regions are determined using a geometric method based on Lucas-Kanade optical flow and the RANSAC algorithm. We evaluate the accuracy of our system using the KITTI dataset, comparing it to various dynamic feature point selection strategies and DynaSLAM. Experiments show that our proposed system demonstrates a reduction in both absolute trajectory error and relative trajectory error, with a maximum reduction of 39% and 30%, respectively, compared to other systems.

An Improved ORB Feature Extraction Algorithm Based on Enhanced Image and Truncated Adaptive Threshold

“…In recent years, scholars have mainly made two improvements to address the illumination variation problem in the ORB algorithm: the first is to improve the calculation method of adaptive thresholds [17]- [19], and the second is to expand the application scope of adaptive thresholds [20], [21].…”

A review of visual SLAM for robotics: evolution, properties, and future applications

“…Furthermore, ORB-SLAM3 furthers its classification to include all three categories: only-visual, visual-inertial, and RGB-D SLAM. This expansion underscores the adaptability and versatility of ORB-SLAM in real-life applications (Zang et al, 2023;Ca et al, 2021;Campos et al, 2021).…”

“…The subsequent phase involves loop closing, process optimization, and selecting similar candidate data in all versions. However, versions 2 and 3 include additional steps such as bundle adjustment welding and map merging ( Mur-Artal and Tardós, 2017a ; Zang et al, 2023 ). The last stage is preparing the output, focusing on creating the final map that includes essential information such as graphs, lines, point mapping, and 2D and 3D maps for use in the SLAM process ( Acosta-Amaya et al, 2023 ).…”

Mapping with Monocular Camera Sensor under Adversarial Illumination for Intelligent Vehicles

“…It also introduces an inertial sensor to minimize the scale variance of monocular-vision-only SLAM. Further improvements are also proposed with sensor adaptability, dynamic environment, online calibration, and pose-graph reuse [ 32 , 33 , 34 , 35 , 36 ].…”