Monocular 3D object detection in autonomous driving scenarios faces critical challenges, including limited feature representation capabilities of convolutional networks, high computational complexity of Transformer-based models, and heavy reliance on LiDAR assistance. To address these issues, this paper proposes a novel monocular 3D object detection algorithm named MonoSPDC, which integrates global-channel attention optimization and multi-view joint prediction. First, a Global-Channel Feature Reconstruction Module is designed to enhance feature representation in heatmaps by fusing global spatial correlations with channel-wise attention mechanisms. Second, a Multi-View Joint Prediction Structure is developed to dynamically integrate 3D corner geometric constraints with depth regression results, significantly reducing depth estimation errors. Experimental results on the KITTI dataset demonstrate that the proposed method achieves a 3D average precision () of 24.87%, outperforming the state-of-the-art methods MonoDTR and MonoDETR by 13% and 5%, respectively, while improving detection speed by over 16%. On the Waymo dataset, it attains an of 67.55%, surpassing CaDDN and MonoCon by 5% and 10%. Notably, MonoSPDC eliminates the dependency on LiDAR-assisted training, substantially reducing model complexity and data requirements, thereby offering a high-precision, lightweight, and cost-effective solution for 3D perception in autonomous driving systems.
QianR, LaiX, LiX. 3D object detection for autonomous driving: a survey[J]. Pattern Recognition, 2022, 130: 108796.
[2]
WangT, ZhuX, PangJ, et al. FCOS3D: fully convolutional one-stage monocular 3D object detection[C]∥Proceedings of the IEEE/CVF International Conference on Computer Vision, Montreal, Canada, 2021: 913-922.
[3]
MousavianA, AnguelovD, FlynnJ, et al. 3D bounding box estimation using deep learning and geometry[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Hawaii, USA, 2017: 7074-7082.
[4]
ReadingC, HarakehA, ChaeJ, et al. Categorical depth distribution network for monocular 3D object detection[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Online, 2021: 8555-8564.
[5]
CarionN, MassaF, SynnaeveG, et al. End-to-end object detection with transformers[C]∥European Conference on Computer Vision, Online, 2020: 213-229.
[6]
WangY, GuiziliniV C, ZhangT, et al. DETR3D: 3D object detection from multi-view images via 3D-to-2D queries[C]∥Conference on Robot Learning, Baltimore, USA, 2022: 180-191.
[7]
HuangK C, WuT H, SuH T, et al. MonoDTR: monocular 3D object detection with depth-aware transformer[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, New Orleans, USA, 2022: 4012-4021.
[8]
PuF, WangY, DengJ, et al. MonoDGP: monocular 3D object detection with decoupled-query and geometry-error priors[J/OL]. [2024-10-25].
[9]
DuanK, BaiS, XieL, et al. CenterNet: keypoint triplets for object detection[C]∥Proceedings of the IEEE/CVF International Conference on Computer Vision, Seoul, Korea (South), 2019: 6569-6578.
[10]
GeigerA, LenzP, UrtasunR. Are we ready for autonomous driving? The KITTI vision benchmark suite[C]∥IEEE Conference on Computer Vision and Pattern Recognition, Providence, USA, 2012: 3354-3361.
[11]
SunP, KretzschmarH, DotiwallaX, et al. Scalability in perception for autonomous driving: Waymo open dataset[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, USA, 2020: 2446-2454.
[12]
YuF, WangD, ShelhamerE, et al. Deep layer aggregation[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, USA, 2018: 2403-2412.
[13]
ZhuX, HuH, LinS, et al. Deformable ConvNets v2: more deformable, better results[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Long Beach, USA, 2019: 9308-9316.
[14]
ZhangY, LuJ, ZhouJ. Objects are different: flexible monocular 3D object detection[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Online, 2021: 3289-3298.
[15]
WuY, HeK. Group normalization[C]∥Proceedings of the European Conference on Computer Vision, Munich, Germany, 2018: 3-19.
[16]
VaswaniA, ShazeerN, ParmarN, et al. Attention is all you need[J]. Advances in Neural Information Processing Systems, 2017, 30: 1-15.
[17]
PhilionJ, FidlerS. Lift, splat, shoot: encoding images from arbitrary camera rigs by implicitly unprojecting to 3D[C]∥Computer Vision-ECCV 2020: 16th European Conference, Glasgow, UK, 2020: 23-28.
[18]
WangT, ZhuX, PangJ, et al. Probabilistic and geometric depth: detecting objects in perspective[C]∥Conference on Robot Learning, PMLR, Baltimore, USA, 2022: 1475-1485.
[19]
LoshchilovI, HutterF. Decoupled weight decay regularization[C]∥International Conference on Learning Representations, New Orleans, USA, 2019: 1-14.
[20]
LiuZ, WuZ, TóthR. SMOKE: single-stage monocular 3D object detection via keypoint estimation[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, Seattle, USA, 2020: 996-997.
[21]
ZhouY, HeY, ZhuH, et al. Monocular 3D object detection: an extrinsic parameter free approach[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Online, 2021: 7556-7566.
[22]
LiuX, XueN, WuT. Learning auxiliary monocular contexts helps monocular 3D object detection[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2022, 36(2): 1810-1818.
[23]
PhilionJ, FidlerS. Lift, splat, shoot: encoding images from arbitrary camera rigs by implicitly unprojecting to 3D[C]∥European Conference on Computer Vision, Glasgow, UK, 2020: 194-210.
[24]
MaX, ZhangY, XuD, et al. Delving into localization errors for monocular 3D object detection[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Online, 2021: 4721-4730.
[25]
WangL, YuC, SalianS, et al. Searching the deployable convolution neural networks for GPUs[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, New Orleans, USA, 2022: 12227-12236.
[26]
ZhangR, QiuH, WangT, et al. MonoDETR: depth-guided transformer for monocular 3D object detection[C]∥Proceedings of the IEEE/CVF International Conference on Computer Vision, Paris, France, 2023: 9155-9166.
[27]
ChongZ, MaX, ZhangH, et al. MonoDistill: learning spatial features for monocular 3D object detection[C]∥International Conference on Learning Representations, Online, 2022: 1-16.