[1]夏桂华,柏隽,李诗洋,等.自监督异构双教师蒸馏的船舶点云目标检测方法研究[J].智能系统学报,2026,21(5):1268-1281.[doi:10.11992/tis.202606023]
XIA Guihua,BAI Jun,LI Shiyang,et al.Research on self-supervised heterogeneous dual-teacher distillation for ship detection in point clouds[J].CAAI transactions on intelligent systems,2026,21(5):1268-1281.[doi:10.11992/tis.202606023]
点击复制
XIA Guihua,BAI Jun,LI Shiyang,et al.Research on self-supervised heterogeneous dual-teacher distillation for ship detection in point clouds[J].CAAI transactions on intelligent systems,2026,21(5):1268-1281.[doi:10.11992/tis.202606023]
自监督异构双教师蒸馏的船舶点云目标检测方法研究
《智能系统学报》[ISSN 1673-4785/CN 23-1538/TP] 卷:
21
期数:
2026年第5期
页码:
1268-1281
栏目:
学术论文—机器感知与模式识别
出版日期:
2026-09-05
- Title:
- Research on self-supervised heterogeneous dual-teacher distillation for ship detection in point clouds
- Keywords:
- unmanned surface vehicle; ship point cloud; object detection; LiDAR; self-supervised learning; dual-teacher distillation; contrastive learning; feature transfer
- 分类号:
- TP242
- 摘要:
- 随着人工智能技术的快速发展，无人艇自主航行与作业技术成为当前学术界的研究热点之一，海上船舶目标检测与环境感知是保障无人艇自主决策与安全航行的关键技术。激光雷达可提供稳定的三维几何信息，使船舶点云目标检测成为海上环境感知的重要技术路径。针对海上船舶点云目标检测中高质量数据集构建困难、远距离目标回波稀疏以及海面动态杂波干扰强等问题，提出了一种自监督异构双教师对比网络指导的蒸馏方法，以缓解船舶点云目标检测网络对船舶点云数据集的依赖，并提升其在稀疏目标表征和复杂海面背景干扰条件下的检测鲁棒性。实验验证基于“海豚1号”智能船采集的海上船舶点云数据集开展，结果表明，该方法较基线方法平均精度均值(mean average precision，mAP)提升6.4百分点，在稀疏点云和强杂波场景下保持较稳定的检测性能，并具备面向无人艇船载感知平台部署与优化的应用潜力。
- Abstract:
- With the rapid development of artificial intelligence, autonomous navigation and operation technologies for unmanned surface vehicle(USV) have become one of the current research focuses in academia. Maritime ship target detection and environmental perception are key technologies for ensuring autonomous decision-making and safe navigation of USVs. LiDAR can provide stable three-dimensional geometric information, which makes ship point-cloud target detection an important technical approach for maritime environmental perception. To address the problems of difficult construction of high-quality datasets, sparse echoes from long-range targets, and strong interference from dynamic sea clutter in maritime ship point-cloud target detection, this study proposed a distillation method guided by a self-supervised heterogeneous dual-teacher contrastive network. The method aims to reduce the dependence of ship point-cloud target detection networks on ship point-cloud datasets and improve their detection robustness under sparse target representation and complex maritime background interference. Experiments were conducted on a maritime ship point-cloud dataset collected by the “Dolphin-1” intelligent vessel. The results show that the proposed method improves the mean average precision (mAP) by 6.4 percentage points compared with the baseline method, maintains relatively stable detection performance in sparse point-cloud and strong-clutter scenes, and has application potential for deployment and optimization on onboard perception platforms of USVs.
- 参考文献/References:
-
[1] LIU Zhixiang, ZHANG Youmin, YU Xiang, et al. Unmanned surface vehicles: an overview of developments and challenges[J]. Annual reviews in control, 2016, 41: 71-93
[2] BAI Xiangen, LI Bohan, XU Xiaofeng, et al. A review of current research and advances in unmanned surface vehicles[J]. Journal of marine science and application, 2022, 21(2): 47-58
[3] QIAO Yuanyuan, YIN Jiaxin, WANG Wei, et al. Survey of deep learning for autonomous surface vehicles in marine environments[J]. IEEE transactions on intelligent transportation systems, 2023, 24(4): 3678-3701
[4] CHENG Yuwei, JIANG Mengxin, ZHU Jiannan, et al. Are we ready for unmanned surface vehicles in inland waterways? the USVInland multisensor dataset and benchmark[J]. IEEE robotics and automation letters, 2021, 6(2): 3964-3970
[5] ZHANG Wei, JIANG Feng, YANG Chifu, et al. Research on unmanned surface vehicles environment perception based on the fusion of vision and lidar[J]. IEEE access, 2021, 9: 63107-63121
[6] 李小毛, 张鑫, 王文涛, 等. 基于3D激光雷达的无人水面艇海上目标检测[J]. 上海大学学报(自然科学版), 2017, 23(1): 27-36 LI Xiaomao, ZHANG Xin, WANG Wentao, et al. 3D lidar-based marine object detection for USV[J]. Journal of Shanghai University (natural science edition), 2017, 23(1): 27-36
[7] 武淑文, 李燕烯, 张少琛, 等. 基于深度学习的3D点云目标检测研究综述[J]. 遥测遥控, 2024, 45(5): 1-18 WU Shuwen, LI Yanxi, ZHANG Shaochen, et al. 3D object detection methods based on point cloud with deep learning: a survey[J]. Journal of telemetry, tracking and command, 2024, 45(5): 1-18
[8] 鲁斌, 孙洋, 杨振宇. 融合体素图注意力的三维目标检测算法[J]. 智能系统学报, 2024, 19(3): 598-609 LU Bin, SUN Yang, YANG Zhenyu. 3D object detection algorithm with voxel graph attention[J]. CAAI transactions on intelligent systems, 2024, 19(3): 598-609
[9] GEIGER A, LENZ P, URTASUN R. Are we ready for autonomous driving? The KITTI vision benchmark suite[C]//2012 IEEE Conference on Computer Vision and Pattern Recognition. Providence: IEEE, 2012: 3354-3361.
[10] CAESAR H, BANKITI V, LANG A H, et al. nuScenes: a multimodal dataset for autonomous driving[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 11618-11628.
[11] LANG A H, VORA S, CAESAR H, et al. PointPillars: fast encoders for object detection from point clouds[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach: IEEE, 2019: 12689-12697.
[12] YAN Yan, MAO Yuxing, LI Bo. SECOND: sparsely embedded convolutional detection[J]. Sensors, 2018, 18(10): 3337
[13] ZHOU Yin, TUZEL O. VoxelNet: end-to-end learning for point cloud based 3D object detection[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 4490-4499.
[14] SHI Shaoshuai, WANG Xiaogang, LI Hongsheng. PointRCNN: 3D object proposal generation and detection from point cloud[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach: IEEE, 2019: 770-779.
[15] SHI Shaoshuai, GUO Chaoxu, JIANG Li, et al. PV-RCNN: point-voxel feature set abstraction for 3D object detection[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 10526-10535.
[16] YANG Zetong, SUN Yanan, LIU Shu, et al. 3DSSD: point-based 3D single stage object detector[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 11037-11045.
[17] YIN Tianwei, ZHOU Xingyi, KRAHENBUHL P. Center-based 3D object detection and tracking[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE, 2021: 11779-11788.
[18] CHEN Yukang, LIU Jianhui, ZHANG Xiangyu, et al. VoxelNeXt: fully sparse VoxelNet for 3D object detection and tracking[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Vancouver: IEEE, 2023: 21674-21683.
[19] ZHANG Gang, CHEN Junnan, GAO Guohuan, et al. SAFDNet: a simple and effective network for fully sparse 3D object detection[C]//2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2024: 14477-14486.
[20] HE Kaiming, CHEN Xinlei, XIE Saining, et al. Masked autoencoders are scalable vision learners[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans: IEEE, 2022: 15979-15988.
[21] XIE Saining, GU Jiatao, GUO Demi, et al. PointContrast: unsupervised pre-training for 3D point cloud understanding[C]//Computer Vision–ECCV 2020. Cham: Springer, 2020: 574-591.
[22] PANG Yatian, WANG Wenxiao, TAY F E H, et al. Masked autoencoders for Point cloud self-supervised learning[C]//Computer Vision–ECCV 2022. Cham: Springer, 2022: 604-621.
[23] ZHANG Renrui, GUO Ziyu, GAO Peng, et al. Point-M2AE: multi-scale masked autoencoders for hierarchical point cloud pre-training[C]//Advances in Neural Information Processing Systems 35. New Orleans: NeurIPS, 2022: 27061-27074.
[24] MIN Chen, XIAO Liang, ZHAO Dawei, et al. Occupancy-MAE: self-supervised pre-training large-scale LiDAR point clouds with masked occupancy autoencoders[J]. IEEE transactions on intelligent vehicles, 2024, 9(7): 5150-5162
[25] HESS G, JAXING J, SVENSSON E, et al. Masked autoencoder for self-supervised pre-training on lidar point clouds[C]//2023 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops. Waikoloa: IEEE, 2023: 350-359.
[26] LIN Z, WANG Y, QI S, et al. BEV-MAE: Bird’s eye view masked autoencoders for point cloud pre-training in autonomous driving scenarios[C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2024, 38(4): 3443-3451. [26] LIN Zhiwei, WANG Yongtao, QI Shengxiang, et al. BEV-MAE: bird’s eye view masked autoencoders for point cloud pre-training in autonomous driving scenarios[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Vancouver: AAAI, 2024: 3204-3212.
[27] TAYEBATI S, TULABANDHULA T, TRIVEDI A R. Sense less, generate more: pre-training LiDAR perception with masked autoencoders for ultra-efficient 3D sensing[EB/OL]. (2024-06-12)[2026-06-12]. https://doi.org/10.48550/arXiv.2406.07833.
[28] NISAR B, WASLANDER S L. PSA-SSL: pose and size-aware self-supervised learning on LiDAR point clouds[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Music City Center: IEEE, 2025: 6402-6411.
[29] ABDELSAMAD M, ULRICH M, GLASER C, et al. Multi-scale neighborhood occupancy masked autoencoder for self-supervised learning in LiDAR point clouds[C]//2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE, 2025: 22234-22243.
[30] ZHANG Linfeng, DONG Runpei, TAI H S, et al. PointDistiller: structured knowledge distillation towards efficient and compact 3D detection[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Vancouver: IEEE, 2023: 21791-21801.
[31] GAMBASHIDZE A, DADUKIN A, GOLYADKIN M, et al. Weak-to-strong 3D object detection with X-ray distillation[C]//2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2024: 15055-15064.
[32] ZHANG Haonan, LIU Longjun, HUANG Yuqi, et al. CaKDP: category-aware knowledge distillation and pruning framework for lightweight 3D object detection[C]//2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2024: 15331-15341.
[33] PONZINI F, ZACCONE R, MARTELLI M. LiDAR target detection and classification for ship situational awareness: a hybrid learning approach[J]. Applied ocean research, 2025, 158: 104552
[34] 姚婷婷, 张波, 柳晓鸣. 特征增强全卷积网络下的船舶检测[J]. 计算机辅助设计与图形学学报, 2022, 34(7): 1028-1036 YAO Tingting, ZHANG Bo, LIU Xiaoming. Feature enhanced fully convolutional network for ship detection[J]. Journal of computer-aided design & computer graphics, 2022, 34(7): 1028-1036
[35] 李忠智, 尹航, 左剑凯, 等. 基于UNet++网络与多边输出融合策略的船舶检测模型[J]. 计算机工程, 2022, 48(4): 276-283 LI Zhongzhi, YIN Hang, ZUO Jiankai, et al. Ship detection model based on UNet++Network and multiple side-output fusion strategy[J]. Computer engineering, 2022, 48(4): 276-283
[36] 袁国文, 张彩霞, 杨阳, 等. 复杂场景下深度表示的SAR船舶目标检测算法[J]. 计算机工程与应用, 2022, 58(2): 289-294 YUAN Guowen, ZHANG Caixia, YANG Yang, et al. SAR target detection algorithm for depth representation in complex scenes[J]. Computer engineering and applications, 2022, 58(2): 289-294
[37] 何芸倩, 夏桂华, 冯鸿超, 等. 自航船模点云数据集的海上船舶检测[J]. 哈尔滨工程大学学报, 2022, 43(8): 1156-1162,1168 HE Yunqian, XIA Guihua, FENG Hongchao, et al. Marine ship detection on the point cloud dataset of autonomous navigation ship models[J]. Journal of Harbin engineering university, 2022, 43(8): 1156-1162,1168
[38] ZHANG Qiuyu, WANG Lipeng, MENG Hao, et al. Ship detection in maritime scenes under adverse weather conditions[J]. Remote sensing, 2024, 16(9): 1567
[39] XIE Yongchang, NANLAL C, LIU Yuanchang. Reliable LiDAR-based ship detection and tracking for Autonomous Surface Vehicles in busy maritime environments[J]. Ocean engineering, 2024, 312: 119288
- 相似文献/References:
-
[1]夏桂华,朱文序,刘浩岩,等.无人艇集群自组织协同围捕控制算法研究[J].智能系统学报,2025,20(1):162.[doi:10.11992/tis.202405025]
XIA Guihua,ZHU Wenxu,LIU Haoyan,et al.Research on collaborative self-organizing surrounding control algorithm of USV swarm[J].CAAI transactions on intelligent systems,2025,20():162.[doi:10.11992/tis.202405025]
备注/Memo
收稿日期:2026-6-12。
基金项目:国家自然科学基金项目(62403159，52171302).
作者简介:夏桂华，教授，博士生导师，主要研究方向为智能船舶与海洋装备、海洋环境感知及人工智能技术。享受国务院政府特殊津贴，获省部级以上奖励10余项，发表学术论文90余篇。E-mail：xiaguihua@hrbeu.edu.cn。;柏隽，博士研究生，主要研究方向为海上船舶点云目标检测、计算机视觉及海上环境感知。E-mail：baijun19920816@hrbeu.edu.cn。;李诗洋，博士后研究员，主要研究方向为海上船舶环境感知、船舶数据驱动建模。E-mail：lishiyang@dlut.edu.cn。
通讯作者:夏桂华. E-mail：xiaguihua@hrbeu.edu.cn
更新日期/Last Update:
2026-09-05