Interior dense 3D reconstruction system with RGB-D camera for complex large scenes

In mainstream visual inertial odometry systems, the method of positional solution by feature point extraction and matching in the image is widely used. But the tracking accuracy of point features is dependent on texture richness in the environment. Although many existing algorithms introduce line features in the front-end to improve the system's environmental adaptability, most of them sacrifice system real-time in exchange for higher positioning accuracy. The extraction and matching of line features often require more time, thus failing to meet the real-time requirements of the system for localization. In this paper, therefore, we propose a fast visual inertial odometry fused with point and line features, which enables the system to maintain a high level of positioning robustness in dim and changing light environments with low time cost. The point-line features visual inertial odometry algorithm is based on adaptive thresholding of EDLines. By adding an adaptive thresholding component to the EDLines algorithm, the robustness of line feature extraction is enhanced to better adapt to changes in ambient lighting. And the time for line feature extraction is significantly reduced. And a line feature matching algorithm based on geometric information and structural similarity is proposed, which enables fast and accurate line features matching. The algorithm is compared with PL-VIO and VINS-MONO algorithms on the EuRoc dataset and real world secens. Many experiments prove that the algorithm has improved in both real-time and accuracy.

Section: Orthonormal Representationmentioning

confidence: 99%

Fast visual inertial odometry with point–line features using adaptive EDLines algorithm

Zhao¹,

Zhang²,

Wei³

2022

“…For equation (5), its equivalent factor graph is given in figure 3. Now, let us consider the error function of planes.…”

Section: 31mentioning

confidence: 99%

“…When it comes to some higher level human-robotenvironment interaction applications such as autonomous grasp of robots or augmented reality (AR), the SLAM systems tend to adopt dense mapping. Lately, the RGB-D-based dense reconstruction of indoor scenes has exhibited high efficiency and related research work is making progress [5]. As one of the representatives, KinectFusion [6] initiates the small-scaled scene-oriented dense mapping by using ideas of TSDF (abbreviates 'truncated signed distance function').…”

Section: Introductionmentioning

confidence: 99%

A point-line-plane primitives fused localization and object-oriented semantic mapping in structural indoor scenes

Xia¹,

Cui²,

Li³

et al. 2022

Visual SLAM (abbreviates ‘simultaneous localization and mapping’) is a promising solution for environment mapping. This study is devoted to a description of a semantically ensembled SLAM framework. For structural indoor scenes, the structured lines and planes can serve as the newly added constraints to improve the positioning accuracy. In this paper, we propose to comprehensively incorporate point-line-plane primitives and construct a tightly coupled camera pose estimator without any environment assumptions. In particular, the maximum number of extracted lines features is numerically determined. We further integrate a lightweight object mapping pipeline with the designed pose estimator. In this pipeline, the leveraging of fitted plane and cuboid landmarks enables an online, CPU-based dense mapping. The tests on ICL-NUIM and TUM benchmark datasets illustrate that, in comparison to ORB-SLAM2, PL-SLAM (Point and Line based SLAM), SP-SLAM (Supposed Plane SLAM) and PLP-SLAM (Point, Line and Plane fused SLAM), our design leads to superior performances in global consistency and system drift elimination. The feature detection and multi-level navigation map reconstruction results are simultaneously provided.

“…However, with the rapid development of computer vision technology, visual cameras, as low-cost and information-rich sensors, have also demonstrated significant potential in the field of SLAM. As a result, camera-based visual SLAM systems [6,7], have become a popular research topic in recent years. At this stage, visual SLAM algorithms focus primarily on how to generate poses and build maps using cameras, without considering the issue of interference from dynamic environments.…”

Section: Introductionmentioning

confidence: 99%

A dynamic object removing 3D reconstruction system based on multi-sensor fusion

Zhao,

Liu,

Pan

et al. 2024

Self Cite

Currently, one of the key technologies for autonomous navigation of unmanned mobile robots is SLAM, which faces many challenges in practical applications. These challenges include a lack of texture, deterioration in sensor performance, and interference from moving objects in dynamic outdoor environments, all of which have an impact on the mapping system. To address these issues, this paper proposes a framework for lidar, vision camera, and inertial navigation data, resulting in fusion and dynamic object removing. The system consists of three sub-modules: the Lidar-Inertial Module (LIM), the Visual-Inertial Module (VIM), and the Dynamic-Object-Removing Module (DORM). LIM and VIM assist each other, with lidar point clouds providing three-dimensional information for the global voxel map and the camera providing pixel-level color information. At the same time, the DORM performs synchronous dynamic object detection to remove dynamic objects from the global map. The system constructs a multi-sensor factor graph using the state and observation models, and the optimal solution is obtained using least squares. Furthermore, this paper employs triangle descriptors and bundle adjustment methods for loop closure detection in order to reduce accumulated errors and maintain consistency. Experimental results demonstrate that the system can perform clean state estimation, dynamic removing and scene reconstruction in a variety of complex scenarios.