基于自适应分割网络的隧道车道线检测

doi:10.3969/j.issn.1674-8484.2025.03.014

摘要/Abstract

摘要：

针对隧道场景中传统图像分割方案存在依赖先验知识、适应性差等问题，提出一种基于自适应分割网络的车道线检测方法。首先设计了一种基于光照特征的子区域规划方法，通过提取光照特征信号来自适应地判断多区域分割必要性并实时给出对应的子区域配置方案；其次，提出一种基于改进Otsu的车道线区域分割方法，各子区域可独立地根据光照程度调节分割阈值，实现对车道线区域的精确分割；最后设计了一种动态感兴趣区域更新方法，根据前一帧检测结果更新当前帧的感兴趣区域（ROI）。结果表明：在复杂光照、低照度、车道线间断等隧道典型场景下，所提出的算法检测准确率达到96.73%，平均每帧处理时间为24.77 ms；该方法在检测准确率、检测效率与鲁棒性均表现出优势，满足实时性的需求。

关键词: 汽车试验, 车道线检测, 图像分割, 自适应分割, Otsu算法, 动态感兴趣区域（ROI）

Abstract:

Aiming to address the issues of reliance on prior knowledge and limited adaptability in traditional image segmentation approaches within tunnel scenarios, a lane line detection method was proposed based on an adaptive segmentation network. Firstly, a sub-region planning method based on illumination characteristics was designed, which adaptively determined the necessity of multi-region segmentation by extracting illumination feature signals and provided the corresponding sub-region configuration scheme in real time. Secondly, a lane line area segmentation method was proposed based on improved Otsu. Each sub-region can independently adjust the segmentation threshold according to the lighting characteristics to achieve precise segmentation of the lane line area. Finally, a dynamic region of interest update method was designed to update the region of interest (ROI) of the current frame based on the detection results of the previous frame. The results show that the detection accuracy of the proposed algorithm reaches 96.73%, and the average processing time per frame is 24.77 ms in typical tunnel scenarios such as complex lighting, low illumination, and discontinuous lane lines, indicating that the proposed method has the advantages in detection accuracy, detection efficiency and robustness, and can meet the needs of real-time performance.

Key words: automobile test, lane detection, image segmentation, adaptive segmentation, Otsu algorithm, dynamic region of Interest (ROI)

中图分类号:

U467.4

王一飞, 李勇杭, 张雅丽, 王畅, 王泰琪, 袁华智. 基于自适应分割网络的隧道车道线检测[J]. 汽车安全与节能学报, 2025, 16(3): 478-486.

WANG Yifei, LI Yonghang, ZHANG Yali, WANG Chang, WANG Taiqi, YUAN Huazhi. Lane detection algorithm based on adaptive segmentation network[J]. Journal of Automotive Safety and Energy, 2025, 16(3): 478-486.

图/表 15

参考文献 23

[1]	Sultana S, Ahmed B, Paul M, et al. Vision-based robust lane detection and tracking in challenging conditions[J]. IEEE Access, 2023, 11: 67938-67955.
[2]	DONG Yongqi, Patil S, van Arem B, et al. A hybrid spatial-temporal deep learning architecture for lane detection[J]. Comput-Aid Civi Infra Engi, 2023, 38(1): 67-86.
[3]	ZHAO Zhiyuan, WANG Qi, LI Xuelong. Deep reinforcement learning based lane detection and localization[J]. Neurocomput, 2020, 413: 328-338.
[4]	李茂月, 吕虹毓, 河香梅, 等. 自动驾驶中周围车辆识别与信息地图构建技术[J]. 汽车安全与节能学报, 2022, 13(1): 131-141.
	LI Maoyue, Lü Hongyu, HE Xiangmei, et al. Surrounding vehicle recognition and information map construction technology in automatic driving[J]. J Autom Safe Energ, 2022, 13(1): 131-141. (in Chinese)
[5]	王荣, 孙亚夫, 宋娟. 自动驾驶车辆道路测试场景评价方法与试验验证[J]. 汽车工程, 2021, 43(4): 620-628.
	WANG Rong, SUN Yafu, SONG Juan. Evaluation method and test verification of road test scenes for autonomous vehicles[J]. Autom Engineering, 2021, 43(4): 620-628. (in Chinese)
[6]	Muthalagu R, Bolimera A, Kalaichelvi V. Lane detection technique based on perspective transformation and histogram analysis for self-driving cars[J]. Comput Elect Engi, 2020, 85: 106653.
[7]	Dinakaran K, Sagayaraj A S, Kabilesh S K, et al. Advanced lane detection technique for structural highway based on computer vision algorithm[J]. Mate Today: Proc, 2021, 45: 2073-2081.
[8]	陈华, 章秋月, 汤文, 等. 夜间环境下的道路图像增强及车道线检测[J]. 机械设计与制造, 2023(10): 126-129, 133.
	CHEN Hua, ZHANG Qiuyue, TANG Wen, et al. Road image enhancement and road lane detection in night environment[J]. Mach Desi Manufact, 2023(10): 126-129,133. (in Chinese)
[9]	王畅, 李勇杭, 张凯超, 等. 基于融合分割和变尺度窗口的车道线距离检测[J]. 中国公路学报, 2023, 36(7): 212-222. doi: 10.19721/j.cnki.1001-7372.2023.07.018
	WANG Chang, LI Yonghang, ZHANG Kaichao, et al. Lane line distance detection based on fusion segmentation and a variable-scale window[J]. China J Highw Transport, 2023, 36(7): 212-222. (in Chinese)
[10]	GONG Jinliang, CHEN Tao, ZHANG Yanfei. Complex lane detection based on dynamic constraint of the double threshold[J]. Multimed Tool Appl, 2021, 80(18): 27095-27113.
[11]	ZOU Qin, JIANG Hanwen, DAI Qiyu, et al. Robust lane detection from continuous driving scenes using deep neural networks[J]. IEEE Trans Vehi Tech, 2019, 69(1): 41-54.
[12]	Van Gansbeke W, Neven D, De Brabandere B, et al. End-to-end lane detection through differentiable least-squares fitting[C]// Proc IEEE/CVF Int’l Conf Comput Visi Workshops, Seoul, Korea (South), 2019: 905-913.
[13]	武志斐, 李守彪. 基于实例分割的车道线检测算法[J]. 汽车工程, 2023, 45(2): 263-272.
	WU Zhipei, LI Shoubiao. Lane detection algorithm based on instance segmentation[J]. Autom Engineering, 2023, 45(2): 263-272. (in Chinese)
[14]	LI Xiang, LI Jun, HU Xiaolin, et al. Line-cnn: End-to-end traffic line detection with line proposal unit[J]. IEEE Trans Intel Transport Syst, 2019, 21(1): 248-258.
[15]	Tabelini L, Berriel R, Paixao T M, et al. Keep your eyes on the lane: Real-time attention-guided lane detection[C]// Proc IEEE/CVF Conf Comput Visi Patt Recogn. 2021: 294-302.
[16]	Tabelini L, Berriel R, Paixao T M, et al. Polylanenet: Lane estimation via deep polynomial regression[C]// 2020 25th Int’l Conf Patt Recogn (ICPR). IEEE, 2021: 6150-6156.
[17]	FENG Zhengyang, GOU Shaohua, TAN Xin, et al. Rethinking efficient lane detection via curve modeling[C]// Proc IEEE/CVF Conf Comput Visi Patt Recognit, New Orleans, LA, USA, 2022: 17062-17070.
[18]	Etemad M, Soares A, Etemad E, et al. SWS: An unsupervised trajectory segmentation algorithm based on change detection with interpolation kernels[J]. GeoInformat, 2021, 25: 269-289.
[19]	Truong C, Oudre L, Vayatis N. Selective review of offline change point detection methods[J]. Sign Proc, 2020, 167: No 107299.
[20]	Otsu N. A threshold selection method from gray-level histograms[J]. IEEE Trans Syst, Man, Cyber, 1979, 9(1): 62-66.
[21]	XU Xiangyang, XU Shengzhou, JIN Lianghai, et al. Characteristic analysis of Otsu threshold and its applications[J]. Patt Recognit Letts, 2011, 32(7): 956-961.
[22]	Pizer S M, Amburn E P, Austin J D, et al. Adaptive histogram equalization and its variations[J]. Comput Visi, Graph, Imag Proc, 1987, 39(3): 355-368.
[23]	代少升, 肖佳伟, 熊昆, 等. 基于机器视觉的车道线精确检测算法[J]. 半导体光电, 2021, 42(6): 940-946.
	DAI Shaosheng, XIAO Jiawei, XIONG Kun, et al. Accurate detection algorithm of lane line based on machine vision[J]. Semicond Optoelect, 2021, 42(6): 940-946. (in Chinese)

检测方法	正确检测车道数	准确率/ %	平均处理时间/ ms
文献[8]	3 681	87.10	45.29
文献[8]	5 628	91.30	45.29
文献[23]	3 681	87.10	43.23
文献[23]	5 826	94.52	43.23
传统图像分割	3 830	90.63	29.97
传统图像分割	5 717	92.75	29.97
本文算法	4 088	96.73	24.77
本文算法	5 951	96.54	24.77

检测方法	正确检测车道数	准确率/ %	平均处理时间/ ms
文献[8]	3 681	87.10	45.29
文献[8]	5 628	91.30	45.29
文献[23]	3 681	87.10	43.23
文献[23]	5 826	94.52	43.23
传统图像分割	3 830	90.63	29.97
传统图像分割	5 717	92.75	29.97
本文算法	4 088	96.73	24.77
本文算法	5 951	96.54	24.77