基于改进OctConv的车道线检测算法研究  

Research on Lane Detection Algorithm Based on Improved OctConv

作　　者：蒙双 陈乐庚[1] 肖晨晨 MENG Shuang;CHEN Le-geng;XIAO Chen-chen(School of Electronic Engineering and Automation,Guilin University of Electronic Technology,Guilin Guangxi 541000,China)

机构地区：[1]桂林电子科技大学电子工程与自动化学院,广西桂林541000

出　　处：《计算机仿真》2021年第5期142-145,218,共5页Computer Simulation

基　　金：国家自然科学基金(51465011)。

摘　　要：针对当前主流车道线检测算法实时性较差和复杂度高的问题,提出一种改进OctConv的车道线检测方法。根据车道线图像特征明显和非车道线图像冗余的特点,采用OctConv将整体图像信息分为高频分量和低频分量进行车道线特征提取,并结合空洞卷积改善OctConv存在的图像信息丢失问题,实现对车道线图像的高性能语义分割。实验数据表明,提出的方法能有效地提升检测性能,具有较好的拓展性。In the paper, an improved OctConv lane detection method was proposed for solving poor real-time performance and high complexity in current mainstream lane detection algorithms. OctConv was used to divide the whole image into high-frequency components and low-frequency components for lane feature extraction according to the distinct features of lane image and the redundancy of non-lane image. Combining with the dilated convolution, the loss of image information in OctConv was improved to achieve high-performance semantics segmentation of lane image. The experimental results show that the presented method can effectively improve performance and has better scalability.

关 键 词：深度学习 空洞卷积 语义分割 卷积方式 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]