水下视觉SLAM分段式光束平差算法

doi:10.13474/j.cnki.11-2246.2024.1102

测绘通报 ›› 2024, Vol. 0 ›› Issue (11): 7-12.doi: 10.13474/j.cnki.11-2246.2024.1102

水下视觉SLAM分段式光束平差算法

白云鹏^1,2,3,4, 徐会希^1,2,3, 吕凤天^1,2,3

1. 中国科学院沈阳自动化研究所机器人学国家重点实验室, 辽宁沈阳 110016;
2. 中国科学院机器人与智能制造创新研究院, 辽宁沈阳 110169;
3. 辽宁省水下机器人重点实验室, 辽宁沈阳 110169;
4. 中国科学院大学, 北京 100049

收稿日期:2024-03-04 发布日期:2024-12-05
通讯作者: 徐会希,E-mail:xhx@sia.cn
作者简介:白云鹏(1999-),男,硕士,研究方向为水下机器人技术。E-mail:baiyunpeng@sia.cn
基金资助:
中科院战略性先导科技专项(A类)(XDA22040103);辽宁省自然科学基金计划面上项目(2022-MS-035)

Segmented bundle adjustment algorithm for underwater vision SLAM

BAI Yunpeng^1,2,3,4, XU Huixi^1,2,3, Lü Fengtian^1,2,3

1. State Key Laboratory of Robotics, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang 110016, China;
2. Institute for Robotics and Intelligent Manufacturing, Chinese Academy of Sciences, Shenyang 110169, China;
3. Key Laboratory of Marine Robotics, Liaoning Province, Shenyang 110169, China;
4. University of Chinese Academy of Sciences, Beijing 100049, China

Received:2024-03-04 Published:2024-12-05

摘要/Abstract

摘要： 自主水下机器人(AUV)采用视觉SLAM系统可以实现近距离精确定位,但是面对水下大规模场景时,后端优化采用的光束平差算法(BA)存在内存不足和计算效率低的问题,对此本文提出了一种改进的分段式BA优化算法。首先采用基于运动模式的分段方法,根据相机的直行运动和转弯运动将轨迹分段后对每个子分段分别进行BA优化;然后各子分段采用动态调整优化权重的求解方法,根据不同子分段的运动模式动态调整优化参数;最后针对BA代价函数的求解,采用改进的列文伯格-马夸尔特(L-M)求解算法,将信赖域定义为可调参数,优化由雅克比矩阵非正定性引发的算法稳定性问题,提高运算效率。在数据集上的试验结果表明,在运行时间较长、环境较恶劣的数据集序列上,本文算法相较于ORB-SLAM3算法有更好的精度,同时全局BA的效率有显著提高。

关键词: 水下视觉SLAM, 后端优化, 光束平差法, 分段式BA, L-M算法

Abstract: Autonomous underwater vehicle (AUV) can achieve close-range accurate positioning by using visual SLAM system,but when facing large-scale underwater scenes,the back-end optimization using bundle adjustment(BA) algorithm has the problems of insufficient memory and low computational efficiency. To solve these problems,an improved segmented BA optimization algorithm is proposed. A segmentation method based on motion pattern is used to segment the trajectory according to the straight motion and turning motion of the camera,and then BA optimization is performed on each sub-segment respectively. Each sub-segment is solved by dynamically adjusting the optimization weight,and the optimization parameters are dynamically adjusted according to the motion patterns of different sub-segments. For the solving of BA cost function,the improved Levenberg-Marquadt(L-M) algorithm is adopted,the trust region is defined as the tunable parameter,which reduces the non-convergence problem caused by the singularity of the Jacobian matrix and improves the operation efficiency. According to the experimental results on the dataset,the proposed algorithm has better accuracy than the ORB-SLAM3 algorithm when it runs for a long time and the environment is harsh,and the efficiency of the global BA is significantly improved.

Key words: underwater visual SLAM, back-end optimization, bundle adjustment, segment BA, L-M algorithm

中图分类号:

P237

白云鹏, 徐会希, 吕凤天. 水下视觉SLAM分段式光束平差算法[J]. 测绘通报, 2024, 0(11): 7-12.

BAI Yunpeng, XU Huixi, Lü Fengtian. Segmented bundle adjustment algorithm for underwater vision SLAM[J]. Bulletin of Surveying and Mapping, 2024, 0(11): 7-12.

参考文献

[1] 李硕, 刘健, 徐会希, 等.我国深海自主水下机器人的研究现状[J].中国科学:信息科学, 2018, 48(9):1152-1164.
[2] 赵洪全, 徐会希.基于单目直接法的水下视觉惯性里程计方法[J].舰船科学技术, 2022, 44(5):65-69.
[3] 施俊屹, 查富生, 孙立宁, 等.移动机器人视觉惯性SLAM研究进展[J].机器人, 2020, 42(6):734-748.
[4] 刘浩敏, 章国锋, 鲍虎军.面向大尺度场景的单目同时定位与地图构建[J].中国科学(信息科学), 2016, 46(12):1748-1761.
[5] ZHANG Guofeng, LIU Haomin, DONG Zilong, et al.Efficient non-consecutive feature tracking for robust structure-from-motion[J].IEEE Transactions on Image Processing, 2016, 25(12):5957-5970.
[6] KAI N, STEEDLY D, DELLAERT F.Out-of-core bundle adjustment for large-scale 3D reconstruction[C]//Proceedings of the 11th International Conference on Computer Vision.Janeiro:IEEE, 2007.
[7] FANG M, POLLOK T, QU C.Merge-SfM:merging partial reconstructions[C]//Proceedings of the 30th British Machine Vision Conference.Cardiff:[s.n.], 2019.
[8] WU C, AGARWAL S, CURLESS B, et al.Multicore bundle adjustment[C]//Proceedings of 2011 Computer Vision & Pattern Recognition.Colorado Springs:IEEE, 2011.
[9] ERIKSSON A, BASTIAN J, CHIN T J, et al.A consensus-based framework for distributed bundle adjustment[C]//Proceedings of 2016 IEEE Computer Vision & Pattern Recognition.Vegas:IEEE, 2016.
[10] LI X, LING H.Hybrid camera pose estimation with online partitioning for SLAM[J].IEEE Robotics and Automation Letters, 2020, 5(2):1453-1460.
[11] ZHOU L, LUO Z, ZHEN M, et al.Stochastic bundle adjustment for efficient and scalable 3D reconstruction[C]//Proceedings of the 16th European Conference on Computer Vision.Glasgow:ECCV, 2020.
[12] SUGER B, TIPALDI G D, SPINELLO L, et al.An approach to solving large-scale SLAM problems with a small memory footprint[C]//Proceedings of 2014 IEEE International Conference on Robotics and Automation.Hong Kong:IEEE, 2014.
[13] 侯荣波, 魏武, 黄婷, 等.基于ORB-SLAM的室内机器人定位和三维稠密地图构建[J].计算机应用, 2017, 37(5):1439-1444.
[14] 张阳, 李海森, 马礼, 等.基于ORB-SLAM2算法的水下机器人实时定位研究[J].测绘通报, 2019(12):1-7.
[15] FERRERA M, CREUZE V, MORAS J, et al.AQUALOC:an underwater dataset for visual-inertial-pressure localization[J].The International Journal of Robotics Research, 2019, 38(14):1549-1559.
[16] BURRI M, NIKOLIC J, GOHL P, et al.The EuRoC micro aerial vehicle datasets[J].International Journal of Robotics Research, 2016, 35(10):1157-1163.

水下视觉SLAM分段式光束平差算法

Segmented bundle adjustment algorithm for underwater vision SLAM

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 1

编辑推荐

Metrics

本文评价