基于多尺度特征融合的双目位姿估计

朱家奇; 李彬; 赵新华

doi:null

您当前的位置：

首页 >

文章列表页 >

基于多尺度特征融合的双目位姿估计

光学测量与感知 | 更新时间：2025-09-22

- 基于多尺度特征融合的双目位姿估计
- Stereo vision pose estimation based on multi-scale features
- 光学技术 2025年51卷第5期页码：610-616
- 作者机构：
  
  1.天津理工大学计算机科学与工程学院，天津 300384
  2.天津理工大学机械工程学院天津市先进机电系统设计与智能控制重点实验室，天津 300384
- 作者简介：
  
  朱家奇（1995— ），男，博士研究生，从事双目视觉、深度学习、位姿估计研究。13110090339@163.com
  李彬（1981— ），男，教授，博士，从事视觉感知、机器人技术研究。 stereovision@stud.tjut.edu.cn
- 基金信息：
  
  国家自然科学基金项目(52375026)
- DOI：
  中图分类号： TP394.1
- 收稿：2025-04-23，
  
  修回：2025-06-13，
  
  纸质出版：2025-09-20
- 稿件说明：
移动端阅览
朱家奇,李彬,赵新华.基于多尺度特征融合的双目位姿估计[J].光学技术,2025,51(05):610-616.

ZHU Jiaqi,LI Bin,ZHAO Xinhua.Stereo vision pose estimation based on multi-scale features[J].OPTICAL TECHNIQUE,2025,51(05):610-616.
朱家奇,李彬,赵新华.基于多尺度特征融合的双目位姿估计[J].光学技术,2025,51(05):610-616. DOI：

ZHU Jiaqi,LI Bin,ZHAO Xinhua.Stereo vision pose estimation based on multi-scale features[J].OPTICAL TECHNIQUE,2025,51(05):610-616. DOI：

摘要

针对病态区域位姿估计精度不足的问题，提出基于多尺度特征融合的双目位姿估计算法。算法通过构建双目视觉-位姿估计端到端联合优化框架，突破传统级联式架构的信息传递瓶颈。首先，设计多尺度成本积模块，通过融合不同感受野下的上下文特征，提升弱纹理与遮挡区域的立体匹配精度；其次，提出金字塔特征筛选融合模块，利用跨层注意力机制动态聚合低层几何细节与高层语义信息，以增强位姿估计的几何一致性；最后，以Linemod数据集为基础，搭建基于双目的Linemod-Stereo数据集，并验证本文算法的有效性。实验结果表明，所提算法相比传统基于RGB-D的位姿估计方法具有竞争性优势。

Abstract

To address the insufficient pose estimation accuracy in degenerate regions

a stereo pose estimation algorithm based on multi-scale feature fusion is proposed. An end-to-end joint optimization framework for stereo vision and pose estimation is established

overcoming the information transmission bottleneck in traditional cascaded architectures. Firstly

a multi-scale cost volume module is designed to enhance stereo matching accuracy in weakly-textured and occluded regions through context feature fusion across varying receptive fields. Additionally

a pyramid screening-feature fusion module is proposed. This module dynamically aggregates low-level geometric details and high-level semantic information. It uses a cross-layer attention mechanism to achieve this aggregation to significantly improve the geometric consistency of pose estimation. Finally

a new binocular vision-based Linemod-Stereo dataset is constructed to validate the algorithm's effectiveness. Experimental results demonstrate that the proposed method outperforms conventional RGB-D-based approaches.

关键词

Keywords

references

庄宏伟 , 吴东岷 , 张宝顺 . 基于深度语义模型的双目位姿估计 [J ] . 光学技术 , 2023 , 49 （ 02 ）: 205 — 210 .

Zhuang Hongwei , Wu Dongmin , Zhang Baoshun . Binocular pose estimation based on depth semantic model [J ] . Optical Technique , 2023 , 49 （ 02 ）: 205 — 210 .

邓鹏 , 唐文涛 , 黄开明 . 基于级联融合网络的密集型抓取位姿检测 [J ] . 现代制造工程 , 2024 ,（ 07 ）: 126 — 134+60 .

Deng Peng , Tang Wentao , Huang Kaiming . Intensive grasp pose detection based on cascaded fusion network [J ] . Modern Manufacturing Engineering , 2024 ,（ 07 ）: 126 — 134+60 .

Papazov C , Haddadin S , Parusel S , et al . Rigid 3D geometry matching for grasping of known objects in cluttered scenes [J ] . International Journal of Robotics Research , 2012 , 31 （ 4 ）: 538 — 553 .

Alvaro C , Dmitry B , Siddhartha S S , et al . Object recognition and full pose registration from a single image for robotic manipulation [C ] // 2009 IEEE International Conference on Robotics and Automation . Kobe, Japan : IEEE , 2009 : 48 — 55 .

Peng S ， Liu Y , Huang Q , et al . PVNet: pixel-wise voting network for 6 DoF pose estimation [C ] // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . Lone Beach, USA : IEEE , 2019 : 4561 — 4570 .

Park K , Patten T , Vincze M . Pix2pose: pixel-wise coordinate regression of objects for 6D pose estimation [C ] // Proceedings of the IEEE/CVF International Conference on Computer Vision . Seoul, Korea : IEEE , 2019 : 7668 — 7677 .

Nguyen V N , Groueix T , Ponimatkin G , et al . Nope: novel object pose estimation from a single image [C ] // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . Seattle WA, USA : IEEE , 2024 : 17923 — 17932 .

Li Y , Mao Y , Bala R , et al . Mrc-net: 6-dof pose estimation with multiscale residual correlation [C ] // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . Seattle WA, USA : IEEE , 2024 : 10476 — 10486 .

王神龙 , 雍宇 , 吴晨睿 . 基于伪孪生神经网络的低纹理工业零件6D位姿估计 [J ] . 电子学报 , 2023 , 51 （ 01 ）: 192 — 201 .

Wang Shenlong , Yong Yu , Wu Chenrui . 6D Pose estimation of low texture industrial parts based on pseudo-siamese neural network [J ] . Acta Electronica Sinica , 2023 , 51 （ 01 ）: 192 — 201 .

Wang C ， Xu D ， Zhu Y ， et al . DenseFusion：6D object pose estimation by iterative dense fusion [C ] // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . Long Beach, USA : IEEE , 2019 : 3343 — 3352 .

Jiang X , Li D , Chen H , et al . Uni6d: a unified cnn framework without projection breakdown for 6d pose estimation [C ] // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . New Orleans, USA : IEEE , 2022 : 11174 — 11184 .

Mo N , Gan W , Yokoya N , et al . Es6d: a computation efficient and symmetry-aware 6d pose regression framework [C ] // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . New Orleans, USA : IEEE , 2022 : 6718 — 6727 .

He Y , Sun W , Huang H , et al . PVN3D: a deep point-wise 3d keypoints voting network for 6dof pose estimation [C ] // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . Seattle, USA : IEEE , 2020 : 11632 — 11641 .

He Y , Huang H , Fan H , et al . Ffb6d: a full flow bidirectional fusion network for 6d pose estimation [C ] // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . Nashville, USA : IEEE , 2021 : 3003 — 3013 .

Wu J , Chen L , Wang S , et al , Geometric-aware dense matching network for 6D pose estimation of objects from RGB-D images [J ] . Pattern Recognition , 2023 , 137 : 109293

Zhou J , Chen K , Xu L , et al . Deep fusion transformer network with weighted vector-wise keypoints voting for robust 6d object pose estimation [C ] // Proceedings of the IEEE/CVF International Conference on Computer Vision . Vancouver, Canada : IEEE , 2023 : 13967 — 13977 .

Sandler M , Howard A , Zhu M , et al . Mobilenetv2: inverted residuals and linear bottlenecks [C ] // Proceedings of The IEEE Conference on Computer Vision and Pattern Recognition . Salt Lake City, USA : IEEE , 2018 : 4510 — 4520 .

Xu G , Wang X , Ding X , et al . Iterative geometry encoding volume for stereo matching [C ] // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . Vancouver, Canada : IEEE , 2023 : 21919 — 21928 .

Hinterstoisser S , Lepetit V , Ilic S , et al . Model based training, detection and pose estimation of texture-less 3d objects in heavily cluttered scenes [C ] // Computer Vision-ACCV 2012: 11th Asian Conference on Computer Vision , Daejeon, Korea : IEEE , 2013 : 548 — 562 .

Chang J R , Chen Y S . Pyramid stereo matching network [C ] // Proceedings of the IEEE conference on computer vision and pattern recognition . Salt Lake City, USA : IEEE , 2018 : 5410 — 5418 .

Guo X , Yang K , Yang W , et al . Group-wise correlation stereo network [C ] // Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . Long Beach, USA : IEEE , 2019 : 3273 — 3282 .

Lipson L , Teed Z , Deng J . Raft-stereo: multilevel recurrent field transforms for stereo matching [C ] // 2021 International Conference on 3D Vision （3DV） . Prague, Czech Republic : IEEE , 2021 : 218 — 227 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

暂无数据