双分支特征自适应融合的车道线检测方法

doi:10.13229/j.cnki.jdxbgxb.20240397

摘要/Abstract

摘要：

为解决深浅层特征直接融合易导致的特征腐蚀和淹没问题，实现复杂环境车道线精准检测，提出了双分支特征自适应融合的车道线检测方法。首先，设计了双分支特征提取网络，提升复杂环境车道线特征提取能力，减少空间细节信息损失；其次，构建了特征自适应融合模块，利用通道注意力与自注意力引导特征选择和融合，自适应地调整融合过程，优化特征图的通道和空间语义信息；此外，改进的并行混合金字塔池化模块更符合道路细长、大跨度特性，多方向捕获远程上下文关系；最后，本文方法在TuSimple、CULane和Curvelanes数据集进行了实验测试，F₁分别达到了96.93%、76.48%和83.21%，实验结果表明：本文方法能有效应对遮挡、阴影等复杂场景车道线检测任务，其性能相对于主流分割类车道线检测方法有显著提升。

关键词: 计算机应用, 车道线检测, 特征自适应融合, 通道注意力, 自注意力, 金字塔池化

Abstract:

In order to solve the problem of feature corrosion and submergence caused by direct fusion of deep and shallow features and realize accurate lane detection in complex environment， a lane detection method based on adaptive fusion of double-branch features was proposed. Firstly， a dual-branch feature extraction network was designed in the method to enhance the feature extraction capability for lane lines in complex environments and reduce the loss of spatial detail information. Secondly， a feature adaptive fusion module was constructed， in which channel attention and self-attention were utilized to guide feature selection and fusion. The fusion process was adaptively adjusted to optimize the channel and spatial semantic information of feature maps. In addition， the improved parallel hybrid pyramid pooling module is more in line with the characteristics of long and narrow roads and captures remote context in multiple directions. Finally， the proposed method was tested on TuSimple， CULane and Curvelanes data sets， and the F₁ reaches 96.93%， 76.48% and 83.21% respectively. The experimental results show that the proposed method can effectively deal with lane line detection tasks in complex scenes such as occlusion and shadow， and its performance is significantly improved compared with the mainstream segmentation lane line detection methods.

Key words: computer application, lane detection, feature adaptive fusion, channel attention, self-attention, pyramid pooling

中图分类号:

TP391.4

邓天民,谢鹏飞,余洋,陈月田. 双分支特征自适应融合的车道线检测方法[J]. 吉林大学学报(工学版), 2025, 55(12): 3840-3851.

Tian-min DENG,Peng-fei XIE,Yang YU,Yue-tian CHEN. Method of lane detection based on adaptive fusion of double branch features[J]. Journal of Jilin University(Engineering and Technology Edition), 2025, 55(12): 3840-3851.

图/表 13

图1

图2

图3

图4

图5

图6

图7

表1

表2

表3

表4

表5

图8

参考文献 29

[1]	Sun T Y, Tsai S J, Chan V, et al. HSI color model based lane-marking detection[C]∥IEEE Intelligent Transportation Systems Conference,Toronto, Canada, 2006: 1168-1172.
[2]	赵颖, 王书茂, 陈兵旗. 基于改进Hough变换的公路车道线快速检测算法[J]. 中国农业大学学报, 2006(3): 104-108.
	Zhao Ying, Wang Shu-mao, Chen Bing-qi. Fast lane detection algorithm based on improved Hough transform[J]. Journal of China Agricultural University, 2006(3):104-108.
[3]	蔡创新, 邹宇, 潘志庚, 等. 基于多特征融合和窗口搜索的新型车道线检测算法[J]. 江苏大学学报: 自然科学版, 2023, 44(4): 386-391.
	Cai Chuang-xin, Zou Yu, Pan Zhi-geng, et al. Novel lane detection algorithm based on multi-feature fusion and windows searching[J]. Journal of Jiangsu University (Natural Science Edition), 2023, 44(4): 386-391.
[4]	杨金鑫, 范英, 谢纯禄. 基于行距离及粒子滤波的车道线识别算法[J]. 江苏大学学报: 自然科学版, 2020, 41(2): 138-142, 198.
	Yang Jin-xin, Fan Ying, Xie Chun-lu. Lane detection algorithm based on row distance and particle filter[J]. Journal of Jiangsu University (Natural Science Edition), 2020, 41(2): 138-142, 198.
[5]	Tabelini L, Berriel R, Paixao T M, et al. Polylanenet: lane estimation via deep polynomial regression[C]∥25th International Conference on Pattern Recognition (ICPR),Milan, Italy, 2021: 6150-6156.
[6]	Liu R, Yuan Z, Liu T, et al. End-to-end lane shape prediction with transformers[C]∥Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision,Waikoloa, USA, 2021: 3693-3701.
[7]	Feng Z, Guo S, Tan X, et al. Rethinking efficient lane detection via c ourve modeling[C]∥Proceedings of the IEEE/CVF Conferencen Computer Vision and Pattern Recognition,New Orleans,USA,2022: 17041-17049.
[8]	Qin Z, Wang H, Li X. Ultra fast structure-aware deep lane detection[C]∥Computer Vision-ECCV : 16th European Conference, Glasgow, UK, 2020: 276-291.
[9]	张云佐, 郑宇鑫, 武存宇, 等. 基于双特征提取网络的复杂环境车道线精准检测[J].吉林大学学报: 工学版, 2024, 54(7): 1894-1902 .
	Zhang Yun-zuo, Zheng Yu-xin, Wu Cun-yu, et al. Accurate detection of lane lines in complex environment based on dual feature extraction network [J]. Journal of Jilin University (Engineering and Technology Edition), 2024,54 (7): 1894-1902 .
[10]	时小虎, 吴佳琦, 吴春国, 等. 基于残差网络的弯道增强车道线检测方法[J]. 吉林大学学报: 工学版, 2023, 53(2): 584-592.
	Shi Xiao-hu, Wu Jia-qi, Wu Chun-guo, et al. Detection method of enhanced lane lines in curves based on residual network [J]. Journal of Jilin University (Engineering and Technology Edition), 2023,53(2):584-592.
[11]	Liu L, Chen X, Zhu S, et al. CondLaneNet: a top-to-down lane detection framework based on conditional convolution[C]∥Proceedings of the IEEE/CVF International Conference on Computer Vision,Montreal, Canada, 2021:3753-3762.
[12]	Pan X, Shi J, Luo P, et al. Spatial as deep: spatial CNN for traffic scene understanding[C]∥Proceedings of the AAAI Conference on Artificial Intelligence,New Orleans, USA,2018:7276-7283.
[13]	Su J, Chen C, Zhang K, et al. Structure guided lane detection[J/OL].[2024-04-05].
[14]	Zheng T, Fang H, Zhang Y, et al. RESA: recurrent feature-shift aggregator for lane detection[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35(4): 3547-3554.
[15]	Hou Y, Ma Z, Liu C, et al. Learning lightweight lane detection CNNs by self attention distillation[C]∥Proceedings of the IEEE/CVF International Conference on Computer Vision,Seoul, Korea (South),2019: 1013-1021.
[16]	Xu H, Wang S, Cai X, et al. Curvelane-nas: unifying lane-sensitive architecture search and adaptive point blending[C]∥Computer Vision-ECCV : 16th European Conference, Glasgow, UK, 2020: 689-704.
[17]	Pan H, Hong Y, Sun W, et al. Deep dual-resolution networks for realtime and accurate semantic segmentation of traffic scenes[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 24(3): 3448-3460.
[18]	Bae W, Yoo J, Chul Ye J. Beyond deep residual learning for image restoration: persistent homology-guided manifold simplification[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, Honolulu, USA, 2017:1141-1149.
[19]	Cheng M, Su J, Li L, et al. A-DFPN: adversarial learning and deformation feature pyramid networks for object detection[C]∥IEEE 5th International Conference on Image, Vision and Computing (ICIVC),Beijing, China, 2020: 11-18.
[20]	Liu S, Qi L, Qin H, et al. Path aggregation network for instance segmentation[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,Salt Lake City, USA, 2018: 8759-8768.
[21]	Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[C]∥NIPS'17: Proceedings of the 31st International Conference on Neural Information Processing Systems,Long Beach, USA,2017: 6000-6010.
[22]	Zhao H, Shi J, Qi X, et al. Pyramid scene parsing network[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,Honolulu, USA, 2017: 6230-6239.
[23]	TuSimple. TuSimple lane detection benchmark[DB/OL]. [2024-04-06].
[24]	Ghafoorian M, Nugteren C, Baka N, et al. EL-GAN: embedding loss driven generative adversarial networks for lane detection[C]∥Proceedings of the European Conference on Computer Vision (ECCV) Workshops,Munich, Germany, 2018: 256-272.
[25]	Yoo S, Lee H S, Myeong H, et al. End-to-end lane marker detection via row-wise classification[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, Seattle, USA, 2020:4335-4343.
[26]	Tabelini L, Berriel R, Paixao T M, et al. Keep your eyes on the lane: real-time attention-guided lane detection[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Nashville, USA, 2021: 294-302.
[27]	Ko Y, Lee Y, Azam S, et al. Key points estimation and point instance segmentation approach for lane detection[J]. IEEE Transactions on Intelligent Transportation Systems, 2021, 23(7): 8949-8958.
[28]	Chen Z, Liu Q, Lian C.PointLaneNet: efficient end-to-end CNNs for accurate real-time lane detection[C]∥IEEE Intelligent Vehicles Symposium (IV), Paris, France, 2019: 2563-2568.
[29]	Qin Z Q, Zhang P Y, Li X, et al. Ultra fast deep lane detection with hybrid anchor driven ordinal classification[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 46(5): 2555-2568.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

数据集	样本量	训练集	测试集	验证集	分辨率	道路类型
TuSimple	6 408	3 268	2 782	358	1 280×720	高速公路
CULane	133 235	88 880	34 680	9 675	1 640×590	城市，高速公路
Curvelanes	150 000	100 000	30 000	20 000	2 560×1 440	城市，高速公路

方法	F₁/%	Acc/%	FP/%	FN/%
SCNN^［12］	95.97	96.53	6.17	1.80
EL-GAN^［24］	96.26	94.90	4.12	3.36
ENet-SAD^［15］	95.07	96.64	6.02	2.05
ERF-E2E^［25］	96.25	96.02	3.21	4.28
UFLD-ResNet34^［8］	88.02	95.86	18.91	3.75
PolyLaneNet^［5］	90.62	93.36	9.42	9.33
LaneATT^［26］	96.77	95.63	3.53	2.92
DBDNet-ResNet34	96.93	96.72	3.12	3.01

方法	综合/%	正常/%	拥挤/%	夜晚/%	无车道/%	阴影/%	箭头/%	眩光/%	弯道/%	交叉路口/个	FPS
SCNN^［12］	71.60	90.60	69.70	66.10	43.40	66.90	84.10	58.50	64.40	1990	7.5
ENet-SAD^［15］	70.80	90.10	68.80	66.00	41.60	65.90	84.00	60.20	65.70	1998	75
ERFNet-E2E^［25］	74.00	91.00	73.10	67.90	46.60	74.10	85.80	64.50	71.90	2022	—
PINet^［27］	74.40	90.30	72.30	67.70	49.80	68.40	83.70	66.30	65.20	1427	25
UFLD-ResNet34^［8］	72.30	90.70	70.20	66.70	44.40	69.30	85.70	59.50	69.50	2037	170
CurveLanes-NAS-M^［16］	73.50	90.20	70.50	68.20	48.80	69.30	85.70	65.90	67.50	2359	—
RESA-ResNet34^［14］	74.50	91.90	72.40	69.80	46.30	72.00	88.10	66.50	68.60	1896	45.5
LaneATT^［26］	75.13	91.17	76.33	69.51	50.16	73.25	87.82	63.75	68.58	1020	129
DBDNet-ResNet34	76.48	92.55	74.97	70.79	49.67	77.79	88.35	66.63	72.35	1358	76.5

方法	F₁/%	Precision/%	Recall/%
SCNN^［5］	65.02	96.53	6.17
ENet-SAD^［8］	50.31	63.60	41.60
PointLaneNet^［28］	78.47	86.33	71.59
UFLDv2-ResNet34^［29］	81.34	81.49	79.44
CurveLanes-NAS-M^［16］	81.80	93.49	72.71
DBDNet-ResNet34	83.21	86.65	80.03

模型	DBDNet	CPFM	CSAFM	PPM	PHPPM	Aux_Head	F₁/%
ResNet34-Seg	—	—	—	—	—	—	71.37
（a）	√	—	—	—	—	—	73.39
（b）	√	√	—	—	—	—	73.86
（c）	√	—	√	—	—	—	74.76
（d）	√	√	√	—	—	—	75.22
（e）	√	√	√	√	—	—	75.51
（f）	√	√	√	—	√	—	75.74
（g）	√	√	√	—	√	√	76.48