An End-to-End Lane Detection Framework Based on Geometry Transform

Genghua Kou; Weida Wang; Chao Yang; Changle Xiang; Ying Li

doi:10.1007/978-981-99-0479-2_227

An End-to-End Lane Detection Framework Based on Geometry Transform

Genghua Kou, Weida Wang, Chao Yang, Changle Xiang, Ying Li^*

^*Corresponding author for this work

School of Mechanical Engineering

Beijing Institute of Technology

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

3D lane line detection plays an important role in Lane-keeping System, Lane-centering Assist, for intelligent vehicles. Most vision-based methods estimating 3D coordinates of lane lines rely on the inverse-perspective transformation, which affected by the road condition. However, This paper proposes a novel lane line detection framework that is immune to changes in terrain. The proposed framework includes an encoder and two decoders. First, image features are extracted by the feature encoder. Then, the duel-decoder architecture ensures the integrity of the semantic information of the lane lines in the initial image. The correlation between the lane lines is generated by the attention mechanism. Finally, the depth decoder’s output is combined through the geometry transform to obtain the 3D lane line directly. The proposed method explicitly handles the lane line occlusion. Experiments show that our framework has good performance in different driving scenarios.

Original language	English
Title of host publication	Proceedings of 2022 International Conference on Autonomous Unmanned Systems, ICAUS 2022
Editors	Wenxing Fu, Mancang Gu, Yifeng Niu
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	2456-2466
Number of pages	11
ISBN (Print)	9789819904785
DOIs	https://doi.org/10.1007/978-981-99-0479-2_227
Publication status	Published - 2023
Event	International Conference on Autonomous Unmanned Systems, ICAUS 2022 - Xi'an, China Duration: 23 Sept 2022 → 25 Sept 2022

Publication series

Name	Lecture Notes in Electrical Engineering
Volume	1010 LNEE
ISSN (Print)	1876-1100
ISSN (Electronic)	1876-1119

Conference

Conference	International Conference on Autonomous Unmanned Systems, ICAUS 2022
Country/Territory	China
City	Xi'an
Period	23/09/22 → 25/09/22

Keywords

3D lane detection
Depth estimation

Access to Document

10.1007/978-981-99-0479-2_227

Cite this

Kou, G., Wang, W., Yang, C., Xiang, C., & Li, Y. (2023). An End-to-End Lane Detection Framework Based on Geometry Transform. In W. Fu, M. Gu, & Y. Niu (Eds.), Proceedings of 2022 International Conference on Autonomous Unmanned Systems, ICAUS 2022 (pp. 2456-2466). (Lecture Notes in Electrical Engineering; Vol. 1010 LNEE). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-981-99-0479-2_227

Kou, Genghua ; Wang, Weida ; Yang, Chao et al. / An End-to-End Lane Detection Framework Based on Geometry Transform. Proceedings of 2022 International Conference on Autonomous Unmanned Systems, ICAUS 2022. editor / Wenxing Fu ; Mancang Gu ; Yifeng Niu. Springer Science and Business Media Deutschland GmbH, 2023. pp. 2456-2466 (Lecture Notes in Electrical Engineering).

@inproceedings{d35d8f82deba468a8e60df39fe8e7397,

title = "An End-to-End Lane Detection Framework Based on Geometry Transform",

abstract = "3D lane line detection plays an important role in Lane-keeping System, Lane-centering Assist, for intelligent vehicles. Most vision-based methods estimating 3D coordinates of lane lines rely on the inverse-perspective transformation, which affected by the road condition. However, This paper proposes a novel lane line detection framework that is immune to changes in terrain. The proposed framework includes an encoder and two decoders. First, image features are extracted by the feature encoder. Then, the duel-decoder architecture ensures the integrity of the semantic information of the lane lines in the initial image. The correlation between the lane lines is generated by the attention mechanism. Finally, the depth decoder{\textquoteright}s output is combined through the geometry transform to obtain the 3D lane line directly. The proposed method explicitly handles the lane line occlusion. Experiments show that our framework has good performance in different driving scenarios.",

keywords = "3D lane detection, Depth estimation",

author = "Genghua Kou and Weida Wang and Chao Yang and Changle Xiang and Ying Li",

note = "Publisher Copyright: {\textcopyright} 2023, Beijing HIWING Sci. and Tech. Info Inst.; International Conference on Autonomous Unmanned Systems, ICAUS 2022 ; Conference date: 23-09-2022 Through 25-09-2022",

year = "2023",

doi = "10.1007/978-981-99-0479-2_227",

language = "English",

isbn = "9789819904785",

series = "Lecture Notes in Electrical Engineering",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "2456--2466",

editor = "Wenxing Fu and Mancang Gu and Yifeng Niu",

booktitle = "Proceedings of 2022 International Conference on Autonomous Unmanned Systems, ICAUS 2022",

address = "Germany",

}

Kou, G, Wang, W , Yang, C, Xiang, C & Li, Y 2023, An End-to-End Lane Detection Framework Based on Geometry Transform. in W Fu, M Gu & Y Niu (eds), Proceedings of 2022 International Conference on Autonomous Unmanned Systems, ICAUS 2022. Lecture Notes in Electrical Engineering, vol. 1010 LNEE, Springer Science and Business Media Deutschland GmbH, pp. 2456-2466, International Conference on Autonomous Unmanned Systems, ICAUS 2022, Xi'an, China, 23/09/22. https://doi.org/10.1007/978-981-99-0479-2_227

An End-to-End Lane Detection Framework Based on Geometry Transform. / Kou, Genghua; Wang, Weida ; Yang, Chao et al.
Proceedings of 2022 International Conference on Autonomous Unmanned Systems, ICAUS 2022. ed. / Wenxing Fu; Mancang Gu; Yifeng Niu. Springer Science and Business Media Deutschland GmbH, 2023. p. 2456-2466 (Lecture Notes in Electrical Engineering; Vol. 1010 LNEE).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - An End-to-End Lane Detection Framework Based on Geometry Transform

AU - Kou, Genghua

AU - Wang, Weida

AU - Yang, Chao

AU - Xiang, Changle

AU - Li, Ying

PY - 2023

Y1 - 2023

N2 - 3D lane line detection plays an important role in Lane-keeping System, Lane-centering Assist, for intelligent vehicles. Most vision-based methods estimating 3D coordinates of lane lines rely on the inverse-perspective transformation, which affected by the road condition. However, This paper proposes a novel lane line detection framework that is immune to changes in terrain. The proposed framework includes an encoder and two decoders. First, image features are extracted by the feature encoder. Then, the duel-decoder architecture ensures the integrity of the semantic information of the lane lines in the initial image. The correlation between the lane lines is generated by the attention mechanism. Finally, the depth decoder’s output is combined through the geometry transform to obtain the 3D lane line directly. The proposed method explicitly handles the lane line occlusion. Experiments show that our framework has good performance in different driving scenarios.

AB - 3D lane line detection plays an important role in Lane-keeping System, Lane-centering Assist, for intelligent vehicles. Most vision-based methods estimating 3D coordinates of lane lines rely on the inverse-perspective transformation, which affected by the road condition. However, This paper proposes a novel lane line detection framework that is immune to changes in terrain. The proposed framework includes an encoder and two decoders. First, image features are extracted by the feature encoder. Then, the duel-decoder architecture ensures the integrity of the semantic information of the lane lines in the initial image. The correlation between the lane lines is generated by the attention mechanism. Finally, the depth decoder’s output is combined through the geometry transform to obtain the 3D lane line directly. The proposed method explicitly handles the lane line occlusion. Experiments show that our framework has good performance in different driving scenarios.

KW - 3D lane detection

KW - Depth estimation

UR - http://www.scopus.com/inward/record.url?scp=85151064260&partnerID=8YFLogxK

U2 - 10.1007/978-981-99-0479-2_227

DO - 10.1007/978-981-99-0479-2_227

M3 - Conference contribution

AN - SCOPUS:85151064260

SN - 9789819904785

T3 - Lecture Notes in Electrical Engineering

SP - 2456

EP - 2466

BT - Proceedings of 2022 International Conference on Autonomous Unmanned Systems, ICAUS 2022

A2 - Fu, Wenxing

A2 - Gu, Mancang

A2 - Niu, Yifeng

PB - Springer Science and Business Media Deutschland GmbH

T2 - International Conference on Autonomous Unmanned Systems, ICAUS 2022

Y2 - 23 September 2022 through 25 September 2022

ER -

Kou G, Wang W , Yang C, Xiang C, Li Y. An End-to-End Lane Detection Framework Based on Geometry Transform. In Fu W, Gu M, Niu Y, editors, Proceedings of 2022 International Conference on Autonomous Unmanned Systems, ICAUS 2022. Springer Science and Business Media Deutschland GmbH. 2023. p. 2456-2466. (Lecture Notes in Electrical Engineering). doi: 10.1007/978-981-99-0479-2_227