[1]寇旗旗,陈飞宇,张华强,等.基于图像边缘相似性的室内自监督单目深度估计[J].智能系统学报,2026,21(3):713-726.[doi:10.11992/tis.202505005]
KOU Qiqi,CHEN Feiyu,ZHANG Huaqiang,et al.Indoor self-supervised monocular depth estimation based on image edges similarity[J].CAAI Transactions on Intelligent Systems,2026,21(3):713-726.[doi:10.11992/tis.202505005]
点击复制
KOU Qiqi,CHEN Feiyu,ZHANG Huaqiang,et al.Indoor self-supervised monocular depth estimation based on image edges similarity[J].CAAI Transactions on Intelligent Systems,2026,21(3):713-726.[doi:10.11992/tis.202505005]
基于图像边缘相似性的室内自监督单目深度估计
《智能系统学报》[ISSN 1673-4785/CN 23-1538/TP] 卷:
21
期数:
2026年第3期
页码:
713-726
栏目:
学术论文—机器感知与模式识别
出版日期:
2026-05-05
- Title:
- Indoor self-supervised monocular depth estimation based on image edges similarity
- Keywords:
- self-supervision; monocular depth estimation; image edges similarity; feature aggregation; pose optimization; indoor scenes; shape priors; contextual consistency
- 分类号:
- TP391.4
- 文献标志码:
- 2026-2-5
- 摘要:
- 本文针对室内单目深度估计中因结构复杂、边缘重叠严重及旋转分量大导致的深度推理不准确问题，提出一种基于图像边缘相似性的自监督深度估计网络模型。1）引入图像边缘相似性损失函数，作为形状先验约束，缓解因遮挡和重叠引起的性能下降；2）设计自适应特征聚合模块，融合多尺度特征并保持上下文一致性，增强弱相关场景的语义关联；3）提出旋转量优化模块，通过加权融合不同路径的向量来细化位姿估计中的旋转分量，降低旋转误差。实验结果表明，该方法在NYU Depth V2与ScanNet数据集上的深度预测精度分别达到82.9%与78.0%，优于现有先进方法，能够恢复出细节丰富、边缘清晰平滑的深度图，有效提升了室内场景的深度估计效果。
- Abstract:
- In this paper, we propose a self-supervised depth estimation network model based on image edge similarity to address the issue of inaccurate depth inference in indoor monocular depth estimation due to complex structures, severe edge overlapping, and large rotational components. First, we introduce an image edge similarity loss function as a shape prior constraint to mitigate performance degradation caused by occlusions and overlaps. Second, we design an adaptive feature aggregation module to fuse multi-scale features while maintaining contextual consistency, thereby enhancing semantic associations in weakly related scenes. Finally, we propose a rotation optimization module that refines the rotational components in pose estimation by weighted fusion of the vectors from different paths, reducing rotation errors. Experimental results show that our method achieves depth prediction accuracies of 82.9% and 78.0% on the NYU Depth V2 and ScanNet datasets, respectively, outperforming existing state-of-the-art methods. The proposed method can recover depth maps with rich details and clear, smooth edges, effectively improving depth estimation in indoor scenes.
- 参考文献/References:
-
[1] 胡海洋, 陈超平, 高天沐, 等. 单/双目深度估计研究进展与应用综述[J]. 红外与激光工程, 2025, 54(7): 35-48 HU Haiyang, CHEN Chaoping, GAO Tianmu, et al. Recent progress in research and applications of monocular and binocular depth estimation[J]. Infrared and laser engineering, 2025, 54(7): 35-48
[2] 李乐, 张茂军, 熊志辉, 等. 基于内容理解的单幅静态街景图像深度估计[J]. 机器人, 2011, 33(2): 174-180 LI Le, ZHANG Maojun, XIONG Zhihui, et al. Depth estimation from a single still image of street scene based on content understanding[J]. Robot, 2011, 33(2): 174-180
[3] 张楠, 程德强, 寇旗旗, 等. 基于随机遮挡和多粒度特征融合的行人重识别[J]. 北京航空航天大学学报, 2023, 49(12): 3511-3519 ZHANG Nan, CHENG Deqiang, KOU Qiqi, et al. Person re-identification based on random occlusion and multi-granularity feature fusion[J]. Journal of Beijing University of Aeronautics and Astronautics, 2023, 49(12): 3511-3519
[4] EIGEN D, PUHRSCH C, FERGUS R. Depth map prediction from a single image using a multi-scale deep network[J]. Advances in neural information processing systems, 2014, 27.
[5] JI Pan, LI Runze, BHANU B, et al. MonoIndoor: towards good practice of self-supervised monocular depth estimation for indoor environments[C]//2021 IEEE/CVF International Conference on Computer Vision. Montreal: IEEE, 2022: 12767-12776.
[6] LI Boying, HUANG Yuan, LIU Zeyu, et al. StructDepth: Leveraging the structural regularities for self-supervised indoor depth estimation[C]//2021 IEEE/CVF International Conference on Computer Vision. Montreal: IEEE, 2022: 12643-12653.
[7] ZHOU Kaichen, BIAN Jiawang, ZHENG Jianqing, et al. Manydepth2: motion-aware self-supervised monocular depth estimation in dynamic scenes[J]. IEEE robotics and automation letters, 2025, 10(7): 6704-6711
[8] 程德强, 范舒铭, 钱建生, 等. 基于坐标感知注意的多帧自监督单目深度估计[J]. 北京航空航天大学学报, 2025, 51(7): 2218-2228 CHENG Deqiang, FAN Shuming, QIAN Jiansheng, et al. Coordinate-aware attention-based multi-frame self-supervised monocular depth estimation[J]. Journal of Beijing University of Aeronautics and Astronautics, 2025, 51(7): 2218-2228
[9] GARG R, B G V K, CARNEIRO G, et al. Unsupervised CNN for single view depth estimation: geometry to the rescue[C]//Computer Vision–ECCV 2016. Cham: Springer, 2016: 740-756.
[10] GODARD C, MAC AODHA O, BROSTOW G J. Unsupervised monocular depth estimation with left-right consistency[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 6602-6611.
[11] ZHOU Tinghui, BROWN M, SNAVELY N, et al. Unsupervised learning of depth and ego-motion from video[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 6612-6619.
[12] YIN Zhichao, SHI Jianping. GeoNet: unsupervised learning of dense depth, optical flow and camera pose[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 1983-1992.
[13] CASSER V, PIRK S, MAHJOURIAN R, et al. Unsupervised monocular depth and ego-motion learning with structure and semantics[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Long Beach: IEEE, 2020: 381-388.
[14] 程德强, 徐帅, 韩成功, 等. 基于视觉注意的自监督单目深度估计[J]. 计算机辅助设计与图形学学报, 2024, 36(12): 1920-1931 CHENG Deqiang, XU Shuai, HAN Chenggong, et al. Visual attention-based self-supervised monocular depth estimation[J]. Journal of computer-aided design & computer graphics, 2024, 36(12): 1920-1931
[15] 柴国强, 薄祥仕, 刘海军, 等. 基于不确定性单目图像自监督场景深度估计[J]. 北京航空航天大学学报, 2024, 50(12): 3780-3787 CHAI Guoqiang, BO Xiangshi, LIU Haijun, et al. Self-supervised scene depth estimation for monocular images based on uncertainty[J]. Journal of Beijing University of Aeronautics and Astronautics, 2024, 50(12): 3780-3787
[16] GODARD C, MAC AODHA O, FIRMAN M, et al. Digging into self-supervised monocular depth estimation[C]//2019 IEEE/CVF International Conference on Computer Vision. Seoul: IEEE, 2019: 3827-3837.
[17] 宋霄罡, 胡浩越, 宁靖宇, 等. 联合语义分割的自监督单目深度估计方法[J]. 计算机研究与发展, 2024, 61(5): 1336-1347 SONG Xiaogang, HU Haoyue, NING Jingyu, et al. Self-supervised monocular depth estimation method for joint semantic segmentation[J]. Journal of computer research and development, 2024, 61(5): 1336-1347
[18] 沙浩, 刘越, 王涌天, 等. 基于二维图像和三维几何约束神经网络的单目室内深度估计方法[J]. 光学学报, 2022, 42(19): 47-57 SHA Hao, LIU Yue, WANG Yongtian, et al. Monocular indoor depth estimation method based on neural networks with constraints on two-dimensional images and three-dimensional geometry[J]. Acta optica sinica, 2022, 42(19): 47-57
[19] 程德强, 张华强, 寇旗旗, 等. 基于层级特征融合的室内自监督单目深度估计[J]. 光学精密工程, 2023, 31(20): 2993-3009 CHENG Deqiang, ZHANG Huaqiang, KOU Qiqi, et al. Indoor self-supervised monocular depth estimation based on level feature fusion[J]. Optics and precision engineering, 2023, 31(20): 2993-3009
[20] 姚广顺, 孙韶媛, 方建安, 等. 基于红外与雷达的夜间无人车场景深度估计[J]. 激光与光电子学进展, 2017, 54(12): 121003 YAO Guangshun, SUN Shaoyuan, FANG Jian’an, et al. Depth estimation of night driverless vehicle scene based on infrared and radar[J]. Laser & optoelectronics progress, 2017, 54(12): 121003
[21] GUO Xiaotong, ZHAO Huijie, SHAO Shuwei, et al. SIM-MultiDepth: self-supervised indoor monocular multi-frame depth estimation based on texture-aware masking[J]. Remote sensing, 2024, 16(12): 2221
[22] CHENG Anqi, YANG Zhiyuan, ZHU Haiyue, et al. GAM-depth: self-supervised indoor depth estimation leveraging a gradient-aware mask and semantic constraints[C]//2024 IEEE International Conference on Robotics and Automation. Yokohama: IEEE, 2024: 5367-5374.
[23] YE Xinchen, OU Yuxiang, WU Biao, et al. Self-supervised monocular depth estimation from videos via adaptive reconstruction constraints[J]. IEEE transactions on circuits and systems for video technology, 2025, 35(3): 2161-2172
[24] 陈莹, 王一良. 基于密集特征融合的无监督单目深度估计[J]. 电子与信息学报, 2021, 43(10): 2976-2984 CHEN Ying, WANG Yiliang. Unsupervised monocular depth estimation based on dense feature fusion[J]. Journal of electronics & information technology, 2021, 43(10): 2976-2984
[25] HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.
[26] SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[EB/OL]. (2014-09-04)[2025-05-13]. https://arxiv.org/abs/1409.1556.
[27] HUANG Gao, LIU Zhuang, VAN DER MAATEN L, et al. Densely connected convolutional networks[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 2261-2269.
[28] WANG Chaoyang, BUENAPOSADA J M, ZHU Rui, et al. Learning depth from monocular videos using direct methods[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 2022-2030.
[29] MARR D, HILDRETH E. Theory of edge detection[J]. Proceedings of the royal society of London series B biological sciences, 1980, 207(1167): 187-217
[30] KOSCHAN A, ABIDI M. Detection and classification of edges in color images[J]. IEEE signal processing magazine, 2005, 22(1): 64-73
[31] SILBERMAN N, HOIEM D, KOHLI P, et al. Indoor segmentation and support inference from RGBD images[C]//Computer Vision–ECCV 2012. Berlin: Springer, 2012: 746-760.
[32] DAI A, CHANG A X, SAVVA M, et al. ScanNet: richly-annotated 3D reconstructions of indoor scenes[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 2432-2443.
[33] FU Huan, GONG Mingming, WANG Chaohui, et al. Deep ordinal regression network for monocular depth estimation[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 2002-2011.
[34] HU Junjie, OZAY M, ZHANG Yan, et al. Revisiting single image depth estimation: toward higher resolution maps with accurate object boundaries[C]//2019 IEEE Winter Conference on Applications of Computer Vision. Waikoloa Village: IEEE, 2019: 1043-1051.
[35] YIN Wei, LIU Yifan, SHEN Chunhua, et al. Enforcing geometric constraints of virtual normal for depth prediction[C]//2019 IEEE/CVF International Conference on Computer Vision. Seoul: IEEE, 2019: 5683-5692.
[36] FAROOQ BHAT S, ALHASHIM I, WONKA P. AdaBins: depth estimation using adaptive bins[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE, 2021: 4008-4017.
[37] NIKLAUS S, MAI Long, YANG Jimei, et al. 3D Ken Burns effect from a single image[J]. ACM transactions on graphics, 2019, 38(6): 1-15
[38] ZHOU Junsheng, WANG Yuwang, QIN Kaihuai, et al. Moving indoor: unsupervised video depth learning in challenging environments[C]//2019 IEEE/CVF International Conference on Computer Vision. Seoul: IEEE, 2019: 8617-8626.
[39] ZHAO Wang, LIU Shaohui, SHU Yezhi, et al. Towards better generalization: joint depth-pose learning without PoseNet[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 9148-9158.
[40] BIAN Jiawang, ZHAN Huangying, WANG Naiyan, et al. Unsupervised scale-consistent depth learning from video[J]. International journal of computer vision, 2021, 129(9): 2548-2564
[41] YU Zehao, JIN Lei, GAO Shenghua. P2net: patch-match and plane-regularization for unsupervised indoor depth estimation[C]//Computer Vision–ECCV 2020. Cham: Springer International Publishing, 2020: 206-222.
[42] BIAN Jiawang, ZHAN Huangying, WANG Naiyan, et al. Unsupervised depth learning in challenging indoor video: weak rectification to rescue [EB/OL]. (2020-06-04)[2025-05-13]. https://arxiv.org/abs/2006.02708v1.
[43] JIANG Hualie, DING Laiyan, HU Junjie, et al. PLNet: plane and line priors for unsupervised indoor depth estimation[C]//2021 International Conference on 3D Vision. London: IEEE, 2021: 741-750.
[44] BIAN Jiawang, ZHAN Huangying, WANG Naiyan, et al. Auto-rectify network for unsupervised indoor depth estimation[J]. IEEE transactions on pattern analysis and machine intelligence, 2022, 44(12): 9802-9813
[45] LYU Chen, HAN Chenggong, CHEN Junhui, et al. TSD-Depth: Using transformers and self-distilling for self-supervised indoor depth estimation[J]. Optik, 2023, 288: 171219
[46] GUO Xiaotong, ZHAO Huijie, SHAO Shuwei, et al. SPDepth: enhancing self-supervised indoor monocular depth estimation via self-propagation[J]. Future Internet, 2024, 16(10): 375
[47] GUO Xiaotong, ZHAO Huijie, SHAO Shuwei, et al. F2Depth: self-supervised indoor monocular depth estimation via optical flow consistency and feature map synthesis[J]. Engineering applications of artificial intelligence, 2024, 133: 108391
[48] WEI Yi, GUO Hengkai, LU Jiwen, et al. Iterative feature matching for self-supervised indoor depth estimation[J]. IEEE transactions on circuits and systems for video technology, 2022, 32(6): 3839-3852
[49] DONG Li, REN Qingji, SHI Jianyang, et al. BiGeoDepth: leveraging bi-geometric priors for unsupervised monocular depth estimation in indoor environments[J]. IEEE transactions on consumer electronics, 2025, 71(2): 2988-2998
[50] WU Haifeng, GU Shuhang, DUAN Lixin, et al. GeoDepth: from point-to-depth to plane-to-depth modeling for self-supervised monocular depth estimation[C]//2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE, 2025: 11525-11535.
[51] HAN Chenggong, LV Chen, HUANG Xiaolin, et al. PRDepth: pose refinement enhancement-based monocular depth estimation for indoor scenes[J]. IEEE transactions on instrumentation and measurement, 2025, 74: 5028216
- 相似文献/References:
-
[1]莫宏伟,孙琪,孙鹏,等.乳腺钼靶肿块自监督预训练迁移检测方法研究[J].智能系统学报,2024,19(5):1082.[doi:10.11992/tis.202304032]
MO Hongwei,SUN Qi,SUN Peng,et al.Self-supervised pretraining detection of mammographic mass targets in breast[J].CAAI Transactions on Intelligent Systems,2024,19():1082.[doi:10.11992/tis.202304032]
备注/Memo
收稿日期:2025-5-13。
基金项目:中央高校基本科研业务费专项资金项目（2024ZDPYCH1001）；国家自然科学基金项目(52204177, 52304182).
作者简介:寇旗旗，副教授，主要研究方向为图像处理、智能检测与模式识别、图像增强与复原，主持国家自然科学基金项目1项，发表学术论文80余篇，获得发明专利授权13项。E-mail：kouqiqi@cumt.edu.cn。;陈飞宇，硕士研究生，主要研究方向为深度估计、图像质量评价。E-mail：TS23060160P31@cumt.edu.cn。;程德强，教授，主要研究方向为智能传感与控制、图像处理与计算机视觉。主持国家自然科学基金项目3项，发表学术论文120余篇，出版专著2部。E-mail：chengdq@cumt.edu.cn。
通讯作者:程德强. E-mail：chengdq@cumt.edu.cn
更新日期/Last Update:
1900-01-01