一种改进的嵌套残差网络车道线检测算法及其应用
An Improved Lane Detection Algorithm Based on Nested Residuals Network and Its'Application
邓世权 1石昀1
作者信息
摘要
车道线检测是自动驾驶中的核心问题之一,针对自动驾驶难以应对真实道路环境中复杂多变性问题,提出了一种基于嵌套结构的残差网络车道线检测模型.首先通过使用该模型对R2U-Net网络结构进行重构,然后利用构建后的深度学习网络对车道线数据集进行学习和检测.该模型以图森公司发布的大规模车道线检测数据集为基础进行了大量的对比实验,结果表明,使用嵌套残差网络结构模型在车道线检测中取得了较高检测效果,检测准确率达到91％,与其他同类模型相比有显著优势.
Abstract
Lane detection is one of the core steps in autopilot.A residual network lane detection model based on nested structure is proposed to deal with the complex and changeable real road environment in this paper.Firstly,the network structure of R2U-Net is reconstructed with the model,and then the deep learning network is used to learn and detect the lane data set.Based on the large-scale lane de-tection data set released by Tucson Company,a large number of comparative experiments are carried out within this model.The results show that the nested residual network structure model achieves high detection effect in lane detection,and the detection accuracy reaches 91％,which has a significant ad-vantage compared with other similar models.
关键词
自动驾驶/循环卷积神经网络/残差网络/车道线检测/U-Net/R2U-NetKey words
Autopilot/recurrent convolutional neural network(RCNN)/residual network/lane de-tection/U-Net/R2U-Net引用本文复制引用
基金项目
黔东南州科技计划项目(黔东南科合J字[2021]41号)
出版年
2024
NETL
NSTL
万方数据