智能车车道识别与图像处理  

Lane Recognition and Image Processing of Intelligent Vehicle

作　　者：徐翔 王琪 高进可 秦海亭 张钰洁 XU Xiang;WANG Qi;GAO Jin-ke;QIN Hai-ting;ZHANG Yu-jie(Jiangsu University of Science and Technology Zhangjiagang Campus Electromechanical and Power Engineering College,Zhangjiagang 215600 China)

机构地区：[1]江苏科技大学张家港校区机电与动力工程学院,江苏张家港215600

出　　处：《自动化技术与应用》2020年第7期91-95,共5页Techniques of Automation and Applications

基　　金：国家重点研发计划(编号2016YFD0700900)。

摘　　要：本文介绍的是一种搭载CMOS摄像头的智能车,该智能车以树莓派为运算核心对采集的图像进行处理。为了简化图像处理难度,解决不同光照的影响,使用大津算法自适应提取灰度图像的阈值,并根据提取到的阈值对图像进行二值化。使用KNN算法对图像进行滤波降噪。通过多项式拟合算法对离散的车道边界进行多项式拟合处理。经过一系列预处理后,可提高车道边界识别的稳定性和准确性。本文还针对常见的车道元素进行处理。处理后,小车可稳定的通过十字弯道和环岛元素。This paper introduces a smart car equipped with CMOS camera,which processes the collected images with raspberry pie as the computing core.In order to simplify the difficulty of image processing and solve the influence of different illumination,OTSU algorithm is used to extract the threshold value of gray image adaptively,and the image is binarized according to the extracted threshold value.KNN algorithm is used to filter and reduce the image noise.Polynomial fitting method is used to fit the discrete lane boundary.After a series of pretreatments,the stability and accuracy of lane boundary recognition can be improved.This paper also deals with common lane elements.After treatment,the car can be stable through the cross-bend and island elements.

关 键 词：大津算法 KNN算法 多项式拟合 环岛处理 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]