Three‑Dimensional Lane Detection Algorithm of Lidar Based on Adaptive Gating and Dual Pathways

Jie Hu; Nan Chen; Wencai Xu; Minjie Chang; Boyuan Xu; Zhanbin Wang; Qixiang Guo

doi:10.3788/CJL230456

Chinese Journal of Lasers, Volume. 50, Issue 22, 2210001(2023)

Three‑Dimensional Lane Detection Algorithm of Lidar Based on Adaptive Gating and Dual Pathways

Jie Hu^1,2,3, Nan Chen^1,2,3, Wencai Xu^1,2,3、*, Minjie Chang^1,2,3, Boyuan Xu^1,2,3, Zhanbin Wang^1,2,3, and Qixiang Guo⁴

Author Affiliations

¹Hubei Key Laboratory of Advanced Technology for Automotive Components, Wuhan University of Technology, Wuhan 430070, Hubei, China

²Hubei Collaborative Innovation Center for Automotive Components Technology, Wuhan University of Technology, Wuhan 430070, Hubei, China

³Hubei Research Center for New Energy & Intelligent Connected Vehicle, Wuhan University of Technology,Wuhan 430070, Hubei, China

⁴Commercial Product R&D Institute, Dongfeng Automobile Co., Ltd., Wuhan 430100, Hubei, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(17)

Fig. 1. Overall architecture of LLDN-AGDP

Download full size

Fig. 2. Schematic diagram of BEV encoder. (a) Schematic diagram of GFPN structure; (b) transformer encoder

Download full size

Fig. 3. Schematic diagram of dual-pathway multi-stage global feature extraction network framework

Download full size

Fig. 4. Schematic diagram of E-MBCONV module

Download full size

Fig. 5. Dual-pathway module and fusion module. (a) Dual-pathway module; (b) fusion module

Download full size

Fig. 6. Schematic diagram of second-order spatial interaction of AMOG module

Download full size

Fig. 7. Detection head of LLDN-AGDP

Download full size

Fig. 8. Dataset distribution

Download full size

Fig. 9. Visualization of detection results on K-Lane test set. (a) Daytime; (b) night; (c) urban; (d) highway; (e) curve; (f) no occlusion; (g) occlusion Occ-4‒6

Download full size

Fig. 10. Visualization results of attention scores of each network. (a) LLDN-AGDP; (b) LLDN-GFC; (c) LLDN-Swin; (d) RLLDN-LC

Download full size

Fig. 11. Hardware deployment of test platform

Download full size

Fig. 12. Detection results of local data set

Download full size

Table 1. Comparison of confidence F₁ scores of different algorithms on K-Lane test set

View table

Table 1. Comparison of confidence F₁ scores of different algorithms on K-Lane test set

Method	Overall / %	Daytime / %	Night / %	Urban / %	Highway / %	Curve / %	No occlusion /%	Occ-4‒6 / %	Frame rate /（frame·s^-1）
RNF-S13	73.2	72.6	74.0	73.1	73.3	70.5	74.9	63.5	13.1
RNF-C13	78.0	77.6	78.5	77.7	78.3	76.0	79.6	69.3	13.0
RNF-D13	72.1	71.3	73.0	71.9	72.3	69.6	74.0	61.9	13.2
LLDN-GFC^［14］	82.1	82.2	82.0	81.7	82.5	78.0	82.9	75.9	11.6
RLLDN-LC^［15］	82.7	82.5	82.9	81.6	84.0	76.1	83.4	79.1	16.4
LLDN-Swin^［25］	82.5	82.6	82.5	81.6	83.7	75.7	83.0	77.4	10.7
LLDN-CAT^［26］	82.7	82.7	82.6	81.7	83.9	77.3	83.5	77.7	11.2
LLDN-AGDP	84.7	84.6	84.7	84.2	85.3	78.9	85.2	79.4	11.8

Table 2. Comparison of classification F₁ scores of different algorithms on K-Lane test set

View table

Table 2. Comparison of classification F₁ scores of different algorithms on K-Lane test set

Method	Overall / %	Daytime/ %	Night / %	Urban / %	Highway / %	Curve / %	No occlusion /%	Occ-4‒6 / %	Frame rate /（frame·s^-1）
RNF-S13	70.5	70.1	71.0	70.4	70.6	68.1	72.3	59.0	13.1
RNF-C13	75.3	75.1	75.5	74.8	76.0	73.1	77.0	65.3	13.0
RNF-D13	68.6	68.3	69.4	68.7	69.0	66.5	70.7	57.6	13.2
LLDN-GFC^［14］	81.1	81.4	80.7	80.6	81.7	76.7	81.9	75.5	11.6
LLDN-Swin^［25］	81.4	81.1	81.8	79.5	82.8	72.3	81.7	77.3	10.7
LLDN-CAT^［26］	81.5	81.5	81.3	80.7	82.7	76.1	82.3	77.1	11.2
LLDN-AGDP	83.6	83.8	83.5	83.0	84.3	78.1	84.3	79.1	11.8

Table 3. Ablation experiment results of LLDN-AGDP
View table
Table 3. Ablation experiment results of LLDN-AGDP
Method DPS GFPN E-MBCONV AMOG Confidence of F₁ /%
（a） 82.1
（b） √ 83.0
（c） √ √ 83.5
（d） √ √ √ 84.1
（e） √ √ √ √ 84.7

Table 4. Ablation experiment results of DPS
View table
Table 4. Ablation experiment results of DPS
Method
N₁=3，N₂=3
（baseline）
N₁=1，N₂=3（test 1） N₁=6，N₂=3（test 2） N₁=3，N₂=0（test 3） N₁=6，N₂=0（test 4） N₁=9，N₂=0（test 5）
Confidence of F₁ /% 84.7 84.4 84.3 83.5 84.1 83.3

Table 5. Ablation experiment results of E-MBCONV, AMOG and GFPN
View table
Table 5. Ablation experiment results of E-MBCONV, AMOG and GFPN
Method E-MBCONV AMOG GFPN
Baseline MBCONV^［33］ Removing ECA Baseline Removing depthwise convolution Baseline FPN^［20］
Confidence of F₁ /% 84.7 84.7 84.4 84.7 84.0 84.7 84.2

Tools

Get Citation

Copy Citation Text

Jie Hu, Nan Chen, Wencai Xu, Minjie Chang, Boyuan Xu, Zhanbin Wang, Qixiang Guo. Three‑Dimensional Lane Detection Algorithm of Lidar Based on Adaptive Gating and Dual Pathways[J]. Chinese Journal of Lasers, 2023, 50(22): 2210001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: remote sensing and sensor

Received: Jan. 12, 2023

Accepted: Mar. 15, 2023

Published Online: Nov. 7, 2023

The Author Email: Xu Wencai (wencaixu_val@163.com)

DOI:10.3788/CJL230456

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Comparison of confidence F1 scores of different algorithms on K-Lane test set

Table 1. Comparison of confidence F1 scores of different algorithms on K-Lane test set

Table 2. Comparison of classification F1 scores of different algorithms on K-Lane test set

Table 2. Comparison of classification F1 scores of different algorithms on K-Lane test set

Table 3. Ablation experiment results of LLDN-AGDP

Table 3. Ablation experiment results of LLDN-AGDP

Table 4. Ablation experiment results of DPS

Table 4. Ablation experiment results of DPS

Table 5. Ablation experiment results of E-MBCONV, AMOG and GFPN

Table 5. Ablation experiment results of E-MBCONV, AMOG and GFPN

Table 1. Comparison of confidence F₁ scores of different algorithms on K-Lane test set

Table 1. Comparison of confidence F₁ scores of different algorithms on K-Lane test set

Table 2. Comparison of classification F₁ scores of different algorithms on K-Lane test set

Table 2. Comparison of classification F₁ scores of different algorithms on K-Lane test set