Vehicle view recognition and 3D wireframe model reconstruction based on geometric topology

doi:10.3785/j.issn.1008-973X.2025.09.010

Journal of ZheJiang University (Engineering Science)

2025, Vol. 59

Issue (9): 1864-1871 DOI: 10.3785/j.issn.1008-973X.2025.09.010

Vehicle view recognition and 3D wireframe model reconstruction based on geometric topology

Qi WU1(

),Bo WANG1,*(

),Huawei WANG1,Li HU1,Baojun LI2

1. School of Automotive and Traffic Engineering, Wuhan University of Science and Technology, Wuhan 430081, China
2. School of Automotive Engineering, Dalian University of Technology, Dalian 116024, China

Download:

HTML

PDF(2613KB) HTML
Export: BibTeX | EndNote (RIS)

Abstract

In the face of complex design tasks, traditional hand-drawn sketches and 3D reconstruction methods often require a huge investment of time, labor, and financial resources. In order to overcome these challenges, a geometric topology-based viewpoint recognition and 3D wireframe model reconstruction method for automobiles was proposed. A model library containing a small number of 3D wireframe models of automobiles was established, which was expanded to 3528 models by multi-scale scaling of body height, width, wheelbase and other parameters. The key point labeling of the car body adopted an interactive system developed based on Matlab, which utilized the labeled key points to match with the projection point sets of different viewpoints in the 3D wireframe model library to identify the angles of the car. The reconstruction of the complex 3D wireframe model was accomplished by using the least squares method with constraints based on the topological relationships of the key points in the 3D model. The reconstruction of 3D wireframe model of arbitrary vehicle types could be realized by updating the model in the model library and the positions of key points. The experimental results showed that the reconstruction time was about 33~44 seconds, and the error of key points was controlled within 32 mm. The reconstruction time and reconstruction error of this method were significantly lower those of the traditional reconstruction method in different viewpoints, demonstrating higher reconstruction efficiency and accuracy.

Key words： car modeling geometric topology key point vehicle angle recognition 3D wireframe reconstruction

Received: 04 September 2023 Published: 25 August 2025

CLC:

TP 393

Fund: 国家自然科学基金资助项目（52375260，51905389）.

Corresponding Authors: Bo WANG E-mail: 2258151731@qq.com;wangbo66@wust.edu.cn

	Service
	E-mail this article
	Add to my bookshelf
	Add to citation manager
	E-mail Alert
	RSS
	Articles by authors
	Qi WU
	Bo WANG
	Huawei WANG
	Li HU
	Baojun LI

Cite this article:

Qi WU,Bo WANG,Huawei WANG,Li HU,Baojun LI. Vehicle view recognition and 3D wireframe model reconstruction based on geometric topology. Journal of ZheJiang University (Engineering Science), 2025, 59(9): 1864-1871.

URL:

https://www.zjujournals.com/eng/10.3785/j.issn.1008-973X.2025.09.010 OR https://www.zjujournals.com/eng/Y2025/V59/I9/1864

基于几何拓扑的汽车视角识别及三维线框模型重建

面对复杂设计任务，传统手绘草图和三维重建方法常常须投入巨大的时间、人力和财力. 为了克服这些挑战，提出基于几何拓扑的汽车视角识别及三维线框模型重建方法. 建立包含少量汽车三维线框模型的模型库，通过对车身高度、宽度、轴距等参数进行多尺度缩放，扩充至3528个模型. 车身关键点标注采用基于Matlab开发的交互系统，利用标注的关键点与三维线框模型库中不同视角的投影点集进行匹配，对汽车角度进行识别. 根据关键点在三维模型中的关联关系，利用带约束的最小二乘法，完成复杂三维线框模型的重建. 通过更新模型库中模型以及关键点位置即可实现对任意车型三维线框模型的重建. 对SUV和MPV两种不同车型的三维线框模型进行重建，实验结果表明，重建结果耗时约33~44 s，关键点误差小于32 mm，该方法在不同视角下的重建时间和重建误差显著低于传统的重建方法，具有更高的重建效率和精度.

关键词： 汽车造型, 几何拓扑, 关键点, 汽车角度识别, 三维线框重建

Fig.1 3D wireframe template

Fig.2 3D wireframe model library

Tab.1 Related data of model library expansion

Fig.3 Key point location diagram

Fig.4 Key point annotation interface

Fig.5 Car body coordinate system diagram

Fig.6 3D shape display at different rotation angles

Fig.7 Key point labeling results from different perspectives

Fig.8 Reconstruction of 3D wireframe models

Fig.9 Reconstruction results under different perspectives

Fig.10 Plotting results of reconstructed wireframe model in image

Fig.11 Comparison of reconstruction error between proposed approach and approach in reference [24]

Fig.12 Showcase of different SUV models in model library

Fig.13 Plotting results of reconstructed wireframe model in image

Fig.14 SUV model key points reconstruction error

Fig.15 Showcase of different MPV models in model library

Fig.16 Plotting results of reconstructed wireframe model in image

Fig.17 Key point reconstruction error of MVP model


[1]	周超, 潘铎现代汽车造型分析与设计[J]. 时代汽车, 2022, (4): 124- 125 ZHOU Chao, PAN Duo Analysis and design of modern automobile styling[J]. Auto Time, 2022, (4): 124- 125 doi: 10.3969/j.issn.1672-9668.2022.04.053

[2]	苏佳幸, 李伽熙, 李睿思, 等基于文本数据的汽车造型需求分析[J]. 时代汽车, 2023, (21): 150- 153 SU Jiaxing, LI Jiaxi, LI Ruisi, et al Analysis of the automotive styling demand based on text data[J]. Auto Time, 2023, (21): 150- 153 doi: 10.3969/j.issn.1672-9668.2023.21.050

[3]	WANG H, YANG J, LIANG W, et al Deep single-view 3D object reconstruction with visual hull embedding[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2019, 33 (1): 8941- 8948 doi: 10.1609/aaai.v33i01.33018941

[4]	LIU F, LIU X. 2D GANs meet unsupervised single-view 3D reconstruction [C]// European Conference on Computer Vision. Cham: Springer Nature Switzerland, 2022: 497–514.

[5]	WALLACE B, HARIHARAN B. Few-shot generalization for single-image 3D reconstruction via priors [C]// Proceedings of the IEEE/CVF International Conference on Computer Vision. Seoul: IEEE, 2019: 3817−3826.

[6]	KATO H, HARADA T. Learning view priors for single-view 3D reconstruction [C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach: IEEE, 2019: 9770−9779.

[7]	WU J, ZHANG C, ZHANG X, et al. Learning shape priors for single-view 3D completion and reconstruction [C]// European Conference on Computer Vision. Cham: Springer International Publishing, 2018: 673–691.

[8]	XU Q, WANG W, CEYLAN D, et al. DISN: Deep implicit surface network for high-quality single-view 3d reconstru- ction[J]. Advances in neural information processing syste- ms , 2019, 32.

[9]	REDDY N D, VO M, NARASIMHAN S G. CarFusion: combining point tracking and part detection for dynamic 3D reconstruction of vehicles [C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 1906–1915.

[10]	TATARCHENKO M, RICHTER S R, RANFTL R, et al. What do single-view 3D reconstruction networks learn? [C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach: IEEE, 2019: 3400-3409.

[11]	HOANG D C, LILIENTHAL A J, STOYANOV T Object-RPE: dense 3D reconstruction and pose estimation with convolutional neural networks[J]. Robotics and Autonomous Systems, 2020, 133: 103632 doi: 10.1016/j.robot.2020.103632

[12]	WANG B, WU Q, WANG H, et al 3D surface reconstruction of car body based on any single view[J]. IEEE Access, 2024, 12: 74903- 74914 doi: 10.1109/ACCESS.2024.3404635

[13]	HOU X, GOU B, CHEN D, et al A method to assist designers in optimizing the exterior styling of vehicles based on key features[J]. Expert Systems with Applications, 2024, 254: 124485 doi: 10.1016/j.eswa.2024.124485

[14]	ROBINSON D E. Fashion theory and product design [M]// Fashion marketing. London: Routledge, 2024: 433–450.

[15]	LU C, KONIUSZ P Detect any keypoints: an efficient light-weight few-shot keypoint detector[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38 (4): 3882- 3890 doi: 10.1609/aaai.v38i4.28180

[16]	CAI G, CHEN B M, LEE T H. Coordinate systems and transformations [M]// Unmanned rotorcraft systems. London: Springer London, 2011: 23–34.

[17]	GOSZTOLAI A, GÜNEL S, LOBATO-RÍOS V, et al LiftPose3D, a deep learning-based approach for transforming two-dimensional to three-dimensional poses in laboratory animals[J]. Nature Methods, 2021, 18 (8): 975- 981 doi: 10.1038/s41592-021-01226-z

[18]	GAO C, SUN C, SHAN L, et al. Rotate3D: representing relations as rotations in three-dimensional space for knowledge graph embedding [C]// Proceedings of the 29th ACM International Conference on Information and Knowledge Management. [s. l. ]: ACM, 2020: 385−394.

[19]	HUANG Y, ZHENG W, ZHANG Y, et al. Tri-perspective view for vision-based 3D semantic occupancy prediction [C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. [S.l.]: IEEE, 2023: 9223–9232.

[20]	NGUYEN T X, BUI H V Study of the influence of lateral forces and velocity on the lateral dynamics of automobile[J]. EUREKA: Physics and Engineering, 2024, (2): 70- 78

[21]	ZHAO R, WANG H, ZHANG C, et al. PointNeuron: 3D neuron reconstruction via geometry and topology learning of point clouds [C]// Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. Waikoloa: IEEE, 2023: 5776–5786.

[22]	WANG Y, WU Y, LI D, et al Millimeter-wave radar and vision fusion-based semantic simultaneous localization and mapping[J]. IEEE Antennas and Wireless Propagation Letters, 2024, 23 (11): 3977- 3981 doi: 10.1109/LAWP.2024.3389678

[23]	DI NUCCI D, SIMONI A, TOMEI M, et al. KRONC: keypoint-based robust camera optimization for 3D car reconstruction [C]// European Conference on Computer Vision. Cham: Springer, 2025: 140−157.

[24]	王博, 江祖毅基于单视图稀疏点的汽车三维模型重建[J]. 武汉科技大学学报, 2023, 46 (4): 296- 302 WANG Bo, JIANG Zuyi Reconstructing 3D automobile model from sparse points on a single view[J]. Journal of Wuhan University of Science and Technology, 2023, 46 (4): 296- 302 doi: 10.3969/j.issn.1674-3644.2023.04.008

[1]	Zhichao ZHU,Jianqiang LI,Hongzhi QI,Qing ZHAO,Qi GAO,Siying LI,Jiayi CAI,Jinyan SHEN. Large model knowledge-guided composite multi-attention method for document-level relation extraction[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(9): 1793-1802.

[2]	Wenxin CHENG,Guanghui YAN,Wenwen CHANG,Baijing WU,Yaning HUANG. Channel-weighted multimodal feature fusion for EEG-based fatigue driving detection[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(9): 1775-1783.

[3]	Xu ZHOU,Ziyu YANG,Junwei ZHANG,Yanling WU,Xiaojie LIN,Wei ZHONG,Baoqin LIU. Optimization of secondary network operation of heating system based on user load quantification[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(8): 1624-1633.

[4]	Hong ZHANG,Xuecheng ZHANG,Guoqiang WANG,Panlong GU,Nan JIANG. Real-time positioning and control of soft robot based on three-dimensional vision[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(8): 1574-1582.

[5]	Hui XIONG,Banglu GE,Jinzhen LIU,Jiaxing WANG. Improved slime mould bee colony algorithm for multi-UAVs cooperative path planning[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(8): 1698-1707.

[6]	Xuanyun LIU,Ying YAN,Zhiyong YU,Fangwan HUANG. Topological design of brain-like reservoir based on scale-free network[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(7): 1385-1393.

[7]	Zenan TIAN,Xinxin GAO,Xiaobin ZHANG. Electrical capacitance tomography imaging of LNG inside pipes using deep neural networks based on flexible sensor[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(6): 1303-1310.

[8]	Xiangyu ZHOU,Yizhi LIU,Yijiang ZHAO,Zhuhua LIAO,Decheng ZHANG. Hierarchical spatial embedding BiGRU model for destination prediction[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(6): 1211-1218.

[9]	Dejun ZHANG,Yanzi BAI,Feng CAO,Yiqi WU,Zhanya XU. Point cloud Transformer adapter for dense prediction task[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(5): 920-928.

[10]	Yu WANG,Chunrong MA,Mingyue ZHAO. Collaborative multi-task assignment of heterogeneous UAVs based on hybrid strategies based multi-objective particle swarm[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(4): 821-831.

[11]	Jiaqi SUN,Yaoling ZHANG,Fei GUO. Reduction of near membrane interface polarization phenomena in membrane distillation through flow channel disturbance[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(4): 863-869.

[12]	Wenqiang CHEN,Dongdan WANG,Wenying ZHU,Yongjie WANG,Tao WANG. Vehicle multimodal trajectory prediction model based on spatio-temporal graph attention network[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(3): 443-450.

[13]	Youping FU,Hang ZHANG,Menghan LI,Jun MENG. Identity recognition based on multi-dimensional features of pulse wave signals[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(3): 566-576.

[14]	Rong HUANG,Wenqin YANG,Zhigang SONG. Blockchain-based trusted distributed industrial data governance scheme[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(2): 269-277.

[15]	Zhuo LI,Jinhu MENG,Shengyan LIN,Yuan GAO,Dawei GU,Chenzhi YOU,Kaihua LIU. Network superspreader measurement algorithm based on multiple filter sketch[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(2): 289-299.

Viewed

Full text

Abstract

Cited

Shared

Discussed