ToF and Stereo Data Fusion Using Dynamic Search Range Stereo Matching

Deng, Yong; Xiao, Jimin; Zhou, Zongtan

doi:10.1109/tmm.2021.3087017

Cited by 7 publications

(6 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The fusion of at least two types of depth-sensing devices has been studied in the past. Notably, the fusion of raw depth maps from two different sensors, such as RGB stereo and time-of-flight (ToF) [70,13,2,21,16,38,3,17], RGB stereo and Lidar [36], RGB and Lidar [55,48,50], RGB stereo and monocular depth [40] and the fusion of multiple RGB stereo algorithms [53] is well-studied and explored. Yet, these methods study specific sensors and are not inherently equipped with 3D reasoning.…”

Section: Multi-sensor Depth Fusionmentioning

confidence: 99%

UncLe-SLAM: Uncertainty Learning for Dense Neural SLAM

Sandström,

Ta,

Van Gool

et al. 2023

2023 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW)

View full text Add to dashboard Cite

We present an uncertainty learning framework for dense neural simultaneous localization and mapping (SLAM). Estimating pixel-wise uncertainties for the depth input of dense SLAM methods allows re-weighing the tracking and mapping losses towards image regions that contain more suitable information that is more reliable for SLAM. To this end, we propose an online framework for sensor uncertainty estimation that can be trained in a self-supervised manner from only 2D input data. We further discuss the advantages of the uncertainty learning for the case of multi-sensor input. Extensive analysis, experimentation, and ablations show that our proposed modeling paradigm improves both mapping and tracking accuracy and often performs better than alternatives that require ground truth depth or 3D. Our experiments show that we achieve a 38% and 27% lower absolute trajectory tracking error (ATE) on the 7-Scenes and TUM-RGBD datasets respectively. On the popular Replica dataset using two types of depth sensors, we report an 11% F1-score improvement on RGBD SLAM compared to the recent state-of-the-art neural implicit approaches.

show abstract

Section: Multi-sensor Depth Fusionmentioning

confidence: 99%

UncLe-SLAM: Uncertainty Learning for Dense Neural SLAM

Sandström,

Ta,

Van Gool

et al. 2023

2023 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW)

View full text Add to dashboard Cite

show abstract

“…On the contrary, pre-fusion is the fusion of Lidar data as a priori information into the stereo vision calculation method to improve the parallax map obtained by stereo vision. This fusion uses the Lidar a priori information to constrain the stereo matching algorithm, the data volume is small and can improve the speed of stereo matching [63][64][65]. Figure 9 shows data fusion flowchart.…”

Section: Lidar + Stereo Visionmentioning

confidence: 99%

“…The structured light is applied to the surface of the object to be measured, and the distorted image of the object surface is captured by two cameras, which are then processed to obtain the information of the corresponding spatial points of the light bars, and finally combined with the intrinsic and extrinsic parameters of the two cameras to recover the 3D spatial information of the object surface projected by the structured light. The structured light binocular vision system consists of two optical cameras, a structured light source, a computer, and other hardware [69]. Structural light binocular visual measurement model is shown in Figure 11.…”

Section: Structure Light + Stereo Visionmentioning

confidence: 99%

Review of key technologies for warehouse 3D reconstruction

Hao

Tan²,

Tai³

et al. 2022

JMES

View full text Add to dashboard Cite

Most of the current warehouse management system is made out of two-dimensional (2D) plane schematic warehouse, which brings a lot of inconvenience to warehouse management, including the warehouse data, storage of goods, location search, inventory, etc. 3D warehouse model began to attract attention as it can provide more intuitive view of warehouse-related information. This paper aims to review and investigate the current key technologies used for 3D modeling of warehouse system. This paper reviewed the method of 3D view reconstruction of the warehouse management system, including the active and passive, active and passive fusion methods, and makes a detailed comparison between the active and passive methods. It was found that different methods were applied to reconstruct the 3D view of warehouse, each with its own advantages and disadvantages.

show abstract

“…In indoor scenarios where firefighting robots perform tasks, GPS signals are often unavailable [5]. The current indoor positioning methods have many methods, including ultra wide band (UWB), inertial measurement unit (IMU), infrared depth sensor (IDS), camera [4,[6][7][8][9][10][11]. UWB has high temporal resolution but is susceptible to indoor environments [4,6].…”

Section: Introductionmentioning

confidence: 99%

A task level fusion autonomous switching mechanism

Lv,

Wang,

Zhang

2023

PLoS ONE

View full text Add to dashboard Cite

Positioning technology is an important component of environmental perception. It is also the basis for autonomous decision-making and motion control of firefighting robots. However, some issues such as positioning in indoor scenarios still remain inherent challenges. The positioning accuracy of the fire emergency reaction dispatching (FERD) system is far from adequate to support some applications for firefighting and rescue in indoor scenarios with multiple obstacles. To solve this problem, this paper proposes a fusion module based on the Blackboard architecture. This module aims to improve the positioning accuracy of a single sensor of the unmanned vehicles within the FERD system. To reduce the risk of autonomous decision-making of the unmanned vehicles, this module uses a comprehensive manner of multiple channels to complement or correct the positioning of the firefighting robots. Specifically, this module has been developed to fusion a variety of relevant processes for precise positioning. This process mainly includes six strategies. These strategies are the denoising, spatial alignment, confidence degree update, observation filtering, data fusion, and fusion decision. These strategies merge with the current scenarios-related parameter data, empirical data on sensor errors, and information to form a series of norms. This paper then proceeds to gain experience data with the confidence degree, error of different sensors, and timeliness of this module by training in an indoor scenario with multiple obstacles. This process is from data of multiple sensors (bottom-level) to control decisions knowledge-based (up-level). This process can obtain globally optimal positioning results. Finally, this paper evaluates the performance of this fusion module for the FERD system. The experimental results show that this fusion module can effectively improve positioning accuracy in an indoor scenario with multiple obstacles. Code is available at https://github.com/lvbingyu-zeze/gopath/tree/master.

show abstract

ToF and Stereo Data Fusion Using Dynamic Search Range Stereo Matching

Cited by 7 publications

References 46 publications

UncLe-SLAM: Uncertainty Learning for Dense Neural SLAM

UncLe-SLAM: Uncertainty Learning for Dense Neural SLAM

Review of key technologies for warehouse 3D reconstruction

A task level fusion autonomous switching mechanism

Contact Info

Product

Resources

About