对采样区地形进行高精度三维重建是保障火星采样作业安全的重要基础。现有的地形重建方法主要依赖双目视觉，但双目视觉视场范围受限、难以覆盖采样机械臂大范围作业区域等问题。与此同时，单目臂载相机具备灵活观测能力，但存在尺度不确定和几何信息不一致等问题。本文开展了基于采样监视双目相机和采样机械臂单目相机的异源图像融合的采样区地形重建研究。利用双目视觉获取具有真实尺度约束的初始地形模型，并通过几何关系估计单目异源相机位姿，将双目深度信息引入单目视角的重建过程，对单目重建结果进行尺度与几何配准，在此基础上实现相对于双目点云的增量式融合制图。实验结果表明，该方法在保持尺度一致性和几何稳定性的前提下，有效扩展了地形重建的空间覆盖范围，提升了复杂作业区域的建模完整性，在相机距离地面1米的条件下，可将单目相机的误差降至毫米级，可为火星样品采集任务中的地形理解与作业评估提供可靠的几何支持。
Abstract: For extraterrestrial exploration missions such as Mars sample return, high-precision and continuously updatable modeling of the sampling site terrain is a critical foundation for autonomous sampling decision-making and operational safety. Existing terrain reconstruction methods mainly rely on stereo vision, which is constrained by limited field of view and baseline configuration, making it difficult to cover the large operational workspace of a sampling manipulator. Although monocular arm-mounted cameras provide flexible viewpoints, they suffer from scale ambiguity and geometric inconsistency. To address these limitations, this study investigates terrain reconstruction under heterogeneous camera collaboration. A multi-camera system composed of a stereo camera and a heterogeneous monocular camera is constructed. Stereo vision is first employed to obtain an initial terrain model with true metric scale. The pose of the heterogeneous camera is then estimated through geometric relationship modeling, enabling the transfer of stereo depth information into the monocular view. By enforcing scale and geometric alignment on the monocular reconstruction, incremental fusion and updating relative to the stereo point cloud are achieved. Experimental results demonstrate that the proposed method effectively extends the spatial coverage of terrain reconstruction while preserving scale consistency and geometric stability, significantly improving modeling completeness in complex operational regions. The approach provides reliable geometric support for terrain understanding and operational assessment in Mars sample collection missions.
[1]KUMAR B M M, ANNAVARAPU R N.Conceptual Design of Mars Sample Return Mission Using Solar Montgolfieres[C]// Aerospace Conference. Montana: IEEE, 2021: 1-10.
[2]ZACNY K, WILSON J, CHU P, et al.Prototype rotary percussive drill for the Mars Sample Return mission[C]// Aerospace Conference. Montana: IEEE, 2011: 1-8.
[3]SCOTT P, DARREN C, DAVID R, et al.The evolution of an orbiting sample container for potential Mars sample return[C]// Aerospace Conference. Montana: IEEE, 2017: 1-16.
[4]郑燕红, 张高, 邓湘金, 等.嫦娥六号月背采样封装系统设计与实现[J].中国科学, 2025, 55(07):1182-1193
[5]LI M, SUN Z, LIU S, et al.Stereo vision technologies for China’s lunar rover exploration mission[J].International Journal of Robotic and Automation, 2016, 31(2):128-136
[6]MA Y, LIU S, WEN B, et al.Weighted Total Least Squares for the Visual Localization of a Planetary Rover[J].Photogrammetric Engineering & Remote Sensing, 2018, 84(10):605-618
[7]史全意.多协作机械臂的视觉定位与世界坐标系标定策略[J].中国科技信息, 2026, 38(2):132-135
[8]YUE Z, WANG Y, LIU L, et al.MC-CIM: A reconfigurable computation-in-memory for efficient stereo matching cost computation [C]// 2022 59th ACM/IEEE Design Automation Conference (DAC). San Francisco, CA, USA: IEEE, 2022: 457-462.
[9]WANG K, BICHOT C E, ZHU C, et al.Pixel to patch sampling structure and local neighboring intensity relationship patterns for texture classification[J].IEEE Signal Processing Letters, 2013, 20(9):853-856
[10]WU Q, LI H, NIU J, et al.Gradient histogram Markov stationary features for image retrieval [C]// IEEE Conference Anthology. China: IEEE, 2013: 1-5.
[11]JO H W, MOON B.A modified census transform using the representative intensity values [C]// 2015 International SoC Design Conference. Gyeongju, South Korea: IEEE, 2015: 309-310.
[12]LI W, YUAN Y, WEI D, et al.Study on engineering cost forecasting of electric power construction based on time response function optimization grey model[C]// 2011 IEEE 3rd International Conference on Communication Software and Networks. Xi' an China: IEEE, 2011: 58-61.
[13]WANG C.A Review on 3D Convolutional Neural Network[C]// 2023 IEEE 3rd International Conference on Power, Electronics and Computer Applications. Shenyang, China: IEEE 2023: 1204-1208.
[14]QU Y, SUN J, TIAN Y, et al.3D Deep Residual Convolutional Neural Network for Underwater Acoustic Source Localization Using Local Acoustic Intensity Field[C]// OCEANS. Chennai, India: IEEE, 2022: 1-5.
[15]SON H, KANG S J.Multi-view stereo with recurrent neural networks for spatio-temporal consistent depth maps[C]// 2023 International Conference on Electronics, Information, and Communication. Singapore: IEEE, 2023: 1-4.
[16]LI R, XUE D, ZHU Y, et al.Self-Supervised Monocular Depth Estimation With Frequency-Based Recurrent Refinement[J].[J].IEEE Transactions on Multimedia, 2023, 25(2):5626-5637
[17]SUN X, ZHAO J, ZHU J, et al.Research and Pilot Application of Visual Detection Method for Phase Distance Between Transformer Bushings Based on Binocular Stereo Vision[C]// 2023 4th International Conference on Information Science, Parallel and Distributed Systems. Guangzhou, China: IEEE, 2023: 364-367.
[18]ZHANG J, FU X, SRIGRAROM S, et al.Depth Estimation in Static Monocular Vision with Stereo Vision Assisted Deep Learning Approach[C]// 2024 4th International Conference on Computer, Control and Robotics. Shanghai, China: IEEE, 2024: 101-107.
[19]XIA Z, WU T, CHEN Z.Binocular Depth Estimation Method for Stereo Matching Based on Pyramid[C]// Transformer 2023 5th International Conference on Robotics, Intelligent Control and Artificial Intelligence. Hangzhou, China: IEEE, 2023: 1066-1069.
[20]Jung J Y, Ho Y S.Depth image interpolation using confidence-based markov random field[J].IEEE Transactions on Consumer Electronics, 2012, 58(4):1399-1402
[21]严亚滔.基于卷积神经网络的复杂场景单目深度估计算法研究[D]. 西安: 西安石油大学, 2025: 30-42. YAN Y. Research on Monocular Depth Estimation Algorithm for Complex Scenes Based on Convolutional Neural Networks [D]. Xi' an: Xi' an Shiyou University, 2025: 30-42.
[22]刘健.基于自监督学习的单目图像深度估计方法研究[D]. 西安: 西安理工大学, 2025: 46-47. LIU J. Research on Monocular Image Depth Estimation Method Based on Self-Supervised Learning [D]. Xi' an: Xi' an University of Technology, 2025: 46-47.
[23]WANG R XU S, DAI C, et al.Moge: Unlocking Accurate Monocular Geometry Estimation for Open-Domain Images with Optimal Training Supervision[C]// Proceedings of the Computer Vision and Pattern Recognition Conference. IEEE, 2025: 5261-5271.
[24]DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al.An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale[C]// International Conference on Learning Representations. 2021.
[25]WEN B, TREPTE M, ARIBIDO J, et al.FoundationStereo: Zero-shot stereo matching[C]// 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville, TN, USA: IEEE, 2025: 5249-5260.
[26]DETONE D, MALISIEWICZ T, RABINOVICH A.SuperPoint: Self-supervised interest point detection and description[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Salt Lake City, UT, USA: IEEE, 2018: 337-33712.
[27]LINDENBERGER P, SARLIN P E, POLLEFEYS M.LightGlue: Local feature matching at light speed[C]// 2023 IEEE/CVF International Conference on Computer Vision. Paris, France: IEEE, 2023: 17581-17592.
[28]RAJU V, PUJARI C.Smart RANSAC: A robust approach[C]// 2024 IEEE 3rd Conference on Information Technology and Data Science. Debrecen, Hungary: IEEE, 2024: 1-5.
[29]UMEYAMA S.Least-squares estimation of transformation parameters between two point patterns[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 1991, 13(4):376-380
CJA