山东科学 ›› 2025, Vol. 38 ›› Issue (1): 129-140.doi: 10.3976/j.issn.1002-4026.20240066

• 交通运输 • 上一篇    

基于注意力机制与动态蛇形卷积的车道线分割算法

宋百玲(), 李星禹(), 刘伟, 邓俊熙, 牟俊麒   

  1. 东北林业大学 机电工程学院,黑龙江 哈尔滨 150006
  • 收稿日期:2024-04-30 出版日期:2025-02-20 发布日期:2025-01-21
  • 通信作者: 李星禹,男,硕士研究生,研究方向为智能驾驶,计算机视觉研究。E-mail:1695352162@qq.com
  • 作者简介:宋百玲(1972—),女,博士,副教授,研究方向为汽车电控系统现代开发(硬件在环、快速控制原型)。E-mail:10559164@qq.com
  • 基金资助:
    黑龙江省“百千万工程”科技重大专项(2021ZX04A01)

Lane segmentation algorithm based on attention mechanism and dynamic snake convolution

SONG Bailing(), LI Xingyu(), LIU Wei, DENG Junxi, MU Junqi   

  1. College of Mechanical and Electrical Engineering,Northeast Forestry University, Harbin 150006, China
  • Received:2024-04-30 Online:2025-02-20 Published:2025-01-21

摘要:

车道检测作为计算机视觉技术在交通领域的关键应用,具有深远的实用价值。尽管如此,现有的语义分割网络模型在道路语义分割任务中仍面临着精度不足和边缘模糊等挑战。针对这些问题,提出了一种基于UNet模型的改进型车道线图像分割网络架构。在UNet模型的跳跃连接中,引入双重注意力机制,优先突出了车道线的重要性,并有效降低了噪声的干扰。此外,采用了动态蛇形卷积来替代传统的卷积方法,增强了网络对车道线的识别能力。考虑到在曝光不足或光线较暗的背景下进行车道线检测的挑战,在图像预处理阶段引入了一种改进的自适应Gamma校正技术,以增强检测的全面性和准确性。同时,在编码器末端引入了空洞金字塔池化技术。实验结果表明,在满足实时性要求的前提下,该模型TuSimple数据集上达到了98.93%的准确率,相较于其他5种基于语义分割的车道线检测算法,展现出更优越的识别效果,结果验证了应用动态蛇形卷积与双注意力机制改进的有效性。

关键词: 车道线检测, 语义分割, 注意力机制, 动态蛇形卷积, Gamma校正算法

Abstract:

Lane detection is a remarkable practical application of computer vision technology in the field of transportation. However, existing semantic segmentation network models still face certain challenges such as insufficient accuracy and blurred edges in road semantic segmentation tasks. To address these issues, an improved lane segmentation network architecture based on the UNet model is proposed. First, a dual attention module (DAM) is introduced in the skip connections of the UNet model, which prioritizes the importance of lane lines and effectively reduces noise interference. Additionally, dynamic snake convolution (DSConv) is employed to replace traditional convolution methods, enhancing the network’s lane detection ability. To enhance the comprehensiveness and accuracy of lane detection in underexposed or dark backgrounds, an improved adaptive Gamma correction method is introduced in the image preprocessing stage. Furthermore, atrous spatial pyramid pooling (ASPP) technology is introduced at the end of the encoder to enhance network performance. Experimental results show that this model achieves an accuracy of 98.93% on the TuSimple dataset while meeting real-time requirements. Compared to five other semantic segmentation-based lane detection algorithms, the proposed algorithm demonstrates superior recognition performance, thus validating its effectiveness.

Key words: lane detection, semantic segmentation, attention mechanism, dynamic snake convolution, Gamma correction algorithm

中图分类号: 

  • U491.6

开放获取 本文遵循知识共享-署名-非商业性4.0国际许可协议(CC BY-NC 4.0),允许第三方对本刊发表的论文自由共享(即在任何媒介以任何形式复制、发行原文)、演绎(即修改、转换或以原文为基础进行创作),必须给出适当的署名,提供指向本文许可协议的链接,同时表明是否对原文作了修改,不得将本文用于商业目的。CC BY-NC 4.0许可协议详情请访问 https://creativecommons.org/licenses/by-nc/4.0