复杂场景下的泊车空间推理模型

doi:10.16182/j.issn1004731x.joss.25-0409

摘要/Abstract

摘要：

针对组合驾驶辅助系统产业化中复杂泊车环境下的车位遮挡、光照不均及检测漏误检问题，提出基于PINet优化的PIPS-Net泊车空间推理模型。在网络架构设计上，将堆叠沙漏网络与循环特征转换聚合器(recurrent feature-shift aggregator，RESA)深度融合，构建上下文特征提取架构以强化对复杂场景的特征推理能力，针对车位检测任务需求重构输出，共同提升复杂场景下车位感知精度；在损失函数设计上，针对车位检测任务特性，设计了针对性损失函数并引入知识蒸馏进行监督，通过多任务损失的协同优化，提升了模型对车位几何结构与状态信息的联合建模能力；s在算法优化方面，提出了基于二维车位空间推理的后处理算法，有效解决车位部分不可见时的检测完整性问题。实验结果表明：该模型表现优异，轻量化版本仍保持高精度检测水平，为自动泊车系统提供了兼具技术先进性与工程实用性的创新解决方案。

关键词: 组合驾驶辅助系统, 自动泊车系统, 泊车空间推理, 循环特征转换聚合器, 知识蒸馏

Abstract:

In the industrialization process of the combined driving assistance system, complex parking environments bring many challenges, such as occlusion of parking spaces, uneven lighting, and missed and false detections. To address these issues, a parking space reasoning model named PIPS-Net was proposed through PINet optimization. In terms of network architecture design, the model deeply integrated the stacked hourglass network with the recurrent feature-shift aggregator (RESA) to construct a context feature extraction architecture, which enhanced the feature reasoning ability in complex scenarios. Meanwhile, it reconstructed the output to meet the requirements of parking space detection tasks, thereby jointly improving the accuracy of parking space perception in complex scenarios. In terms of loss function design,based on the characteristics of parking space detection tasks, targeted loss functions were innovatively designed, and knowledge distillation was introduced for supervision. Through the collaborative optimization of multi-task losses, the model's ability to jointly model the geometric structure and state information of parking spaces was enhanced. In terms of algorithm optimization, a post-processing algorithm based on two-dimensional parking space reasoning was proposed, which effectively solved the problem of detection integrity when parts of the parking space are invisible. Experimental results show that the model performs excellently, and its lightweight version still maintains a high level of detection accuracy, providing an innovative solution with both technical advancement and engineering practicality for automatic parking systems.

Key words: combined driving assistance system, automatic parking system, parking space reasoning, recurrent feature-shift aggregator, knowledge distillation

中图分类号:

TP391.9

周聪玲,王春鹏,谢启伟等 . 复杂场景下的泊车空间推理模型[J]. 系统仿真学报, 2025, 37(11): 2724-2740.

Zhou Congling,Wang Chunpeng,Xie Qiwei,et al . Parking Space Reasoning Model for Complex Scenarios[J]. Journal of System Simulation, 2025, 37(11): 2724-2740.

图/表 20

图1

图2

图3

图4

图5

图6

图7

表1

图8

图9

表2

表3

表4

表5

表6

表7

图10

图11

表8

表9

参考文献 29

[1]	国家市场监督管理总局, 国家标准化管理委员会. 智能网联汽车组合驾驶辅助系统技术要求及试验方法第1部分: 单车道行驶控制: [S]. 北京: 中国标准出版社, 2024: 8-9.
	State Administration for Market Regulation, National Standardization Administration. Intelligent and Connected Vehicle-technical Requirements and Testing Methods for Combined Driver Assistance System-part 1: Single-lane Manoeuvre: [S]. Beijing: Standards Press of China, 2024: 8-9.
[2]	Xie Qiwei, Long Qian, Li Jianping, et al. Application of Intelligence Binocular Vision Sensor: Mobility Solutions for Automotive Perception System[J]. IEEE Sensors Journal, 2024, 24(5): 5578-5592.
[3]	Xie Qiwei, Hu Xiyuan, Ren Lei, et al. A Binocular Vision Application in IoT: Realtime Trustworthy Road Condition Detection System in Passable Area[J]. IEEE Transactions on Industrial Informatics, 2023, 19(1): 973-983.
[4]	王若萱, 吴建平, 徐辉. 自动驾驶汽车感知系统仿真的研究及应用综述[J]. 系统仿真学报, 2022, 34(12): 2507-2521.
	Wang Ruoxuan, Wu Jianping, Xu Hui. Overview of Research and Application on Autonomous Vehicle Oriented Perception System Simulation[J]. Journal of System Simulation, 2022, 34(12): 2507-2521.
[5]	王檀彬, 陈无畏, 李进, 等. 多传感器融合的视觉导航智能车避障仿真研究[J]. 系统仿真学报, 2009, 21(4): 1015-1019.
	Wang Tanbin, Chen Wuwei, Li Jin, et al. Simulations of Obstacle Avoidance for Intelligent Vehicles Based on Vision Navigation and Multi-sensor Fusion[J]. Journal of System Simulation, 2009, 21(4): 1015-1019.
[6]	叶继华, 时淑霞, 李汉曦, 等. 一种结合深度学习和随机森林的地平线检测方法[J]. 系统仿真学报, 2018, 30(7): 2507-2514.
	Ye Jihua, Shi Shuxia, Li Hanxi, et al. A Horizon Detection Method Based on Deep Learning and Random Forest[J]. Journal of System Simulation, 2018, 30(7): 2507-2514.
[7]	蒋弘毅, 王永娟, 康锦煜. 目标检测模型及其优化方法综述[J]. 自动化学报, 2021, 47(6): 1232-1255.
	Jiang Hongyi, Wang Yongjuan, Kang Jinyu. A Survey of Object Detection Models and Its Optimization Methods[J]. Acta Automatica Sinica, 2021, 47(6): 1232-1255.
[8]	罗会兰, 张云. 基于深度网络的图像语义分割综述[J]. 电子学报, 2019, 47(10): 2211-2220.
	Luo Huilan, Zhang Yun. A Survey of Image Semantic Segmentation Based on Deep Network[J]. Acta Electronica Sinica, 2019, 47(10): 2211-2220.
[9]	Zhang Lin, Huang Junhao, Li Xiyuan, et al. Vision-based Parking-slot Detection: a DCNN-based Approach and a Large-scale Benchmark Dataset[J]. IEEE Transactions on Image Processing, 2018, 27(11): 5350-5364.
[10]	Zinelli Andrea, Musto Luigi, Pizzati Fabio. A Deep-learning Approach for Parking Slot Detection on Surround-view Images[C]//2019 IEEE Intelligent Vehicles Symposium (IV). Piscataway: IEEE, 2019: 683-688.
[11]	Zheng Ruitao, Lian Shikang, Liang Weihao, et al. Center Keypoint for Parking Slot Detection with Self-calibrated Convolutions Network[C]//2022 17th International Conference on Control, Automation, Robotics and Vision (ICARCV). Piscataway: IEEE, 2022: 305-310.
[12]	张宝军, 王钊璇, 李孰非. 基于改进卷积神经网络的车位角检测算法研究[J]. 传感器与微系统, 2023, 42(6): 62-65, 69.
	Zhang Baojun, Wang Zhaoxuan, Li Shufei. Research on Parking Angle Detection Algorithm Based on Improved CNN[J]. Transducer and Microsystem Technologies, 2023, 42(6): 62-65, 69.
[13]	Jae Kyu Suhr, Gi Jung Ho. End-to-end Trainable One-stage Parking Slot Detection Integrating Global and Local Information[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(5): 4570-4582.
[14]	Wang Yinan, Guan Yingzhou, Cao Rongchuan. DetPS: A Fully Convolutional End-to-end Parking Slot Detector[C]//2022 IEEE 17th Conference on Industrial Electronics and Applications (ICIEA). Piscataway: IEEE, 2022: 1051-1056.
[15]	Yu Zhuoping, Gao Zhong, Chen Hansheng, et al. SPFCN: Select and Prune the Fully Convolutional Networks for Real-time Parking Slot Detection[C]//2020 IEEE Intelligent Vehicles Symposium (IV). Piscataway: IEEE, 2020: 445-450.
[16]	Jang Chulhoon, Sunwoo Myoungho. Semantic Segmentation-based Parking Space Detection with Standalone Around View Monitoring System[J]. Machine Vision and Applications, 2019, 30(2): 309-319.
[17]	Jiang Wei, Wu Yan, Guan Linting, et al. DFNet: Semantic Segmentation on Panoramic Images with Dynamic Loss Weights and Residual Fusion Block[C]//2019 International Conference on Robotics and Automation (ICRA). Piscataway: IEEE, 2019: 5887-5892.
[18]	Jian Dehui, Lin Changhong. Vision-based Parking Slot Detection Based on End-to-end Semantic Segmentation Training[C]//2020 IEEE International Conference on Consumer Electronics (ICCE). Piscataway: IEEE, 2020: 1-4.
[19]	Lai Cong, Yang Qingyu, Guo Yixin, et al. Semantic Segmentation of Panoramic Images for Real-time Parking Slot Detection[J]. Remote Sensing, 2022, 14(16): 3874.
[20]	Huang Junhao, Zhang Lin, Shen Ying, et al. DMPR-PS: A Novel Approach for Parking-slot Detection Using Directional Marking-point Regression[C]//2019 IEEE International Conference on Multimedia and Expo (ICME). Piscataway: IEEE, 2019: 212-217.
[21]	Min Chen, Xu Jiaolong, Xiao Liang, et al. Attentional Graph Neural Network for Parking-slot Detection[J]. IEEE Robotics and Automation Letters, 2021, 6(2): 3445-3450.
[22]	Ko Yeongmin, Lee Younkwan, Azam Shoaib, et al. Key Points Estimation and Point Instance Segmentation Approach for Lane Detection[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(7): 8949-8958.
[23]	Newell A, Yang Kaiyu, Deng Jia. Stacked Hourglass Networks for Human Pose Estimation[C]//Computer Vision – ECCV 2016. Cham: Springer International Publishing, 2016: 483-499.
[24]	Zheng Tu, Fang Hao, Zhang Yi, et al. RESA: Recurrent Feature-shift Aggregator for Lane Detection[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Palo Alto: AAAI Press, 2021: 3547-3554.
[25]	Sunkara R, Luo Tie. No More Strided Convolutions or Pooling: A New CNN Building Block for Low-resolution Images and Small Objects[C]//Proceedings of the Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Cham: Springer Nature Switzerland, 2023: 443-459.
[26]	Yang Zhendong, Li Zhe, Jiang Xiaohu, et al. Focal and Global Knowledge Distillation for Detectors[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway: IEEE, 2022: 4633-4642.
[27]	Hu Jie, Shen Li, Sun Gang. Squeeze-and-Excitation Networks[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 7132-7141.
[28]	Lin T Y, Goyal P, Girshick R, et al. Focal Loss for Dense Object Detection[C]//2017 IEEE International Conference on Computer Vision (ICCV). Piscataway: IEEE, 2017: 2999-3007.
[29]	Goutte Cyril, Gaussier Eric. A Probabilistic Interpretation of Precision, Recall and F-score, with Implication for Evaluation[C]//Advances in Information Retrieval. Berlin, Heidelberg: Springer Berlin Heidelberg, 2005: 345-359.

名称	版本
操作系统	Ubuntu22.04
CPU	AMD Ryzen™ 7 5800H CPU @3.20 GHz
GPU	NVIDIA® GeForce RTX™ 3050
Python	3.8.20
Pytorch	2.4.1
CUDA	11.8

模块	精确率	召回率	F1分数
RESA+4H	95.90	92.70	94.27
4H	93.12	90.07	91.89
3H	92.15	89.00	90.58
2H	90.34	88.45	89.38
1H	90.28	88.21	89.23
D*1H	95.47	91.80	93.60

模型	精确率	召回率	F1分数
SPFCN^[15]	98.01	97.31	97.65
DMPR-PS^[20]	99.42	99.37	99.39
GCNN^[21]	99.56	99.42	99.49
PIPS-Net	99.46	99.90	99.68
D*PIPS-Net	99.31	99.53	99.41

模型	精确率	召回率	F1分数
SPFCN^[15]	95.09	81.36	87.69
PIPS-Net	95.90	92.70	94.27
D*PIPS-Net	95.47	91.80	93.60

标准差	SPFCN^[15]	PIPS-Net	D*PIPS-Net
σ=0.00	87.69	94.27	93.60
σ=0.50	87.01	94.22	93.05
σ=1.00	86.23	93.05	92.38
σ=1.50	85.57	92.41	91.76