一种基于先验知识的车道识别偏离检测方法  

A method of lane recognition deviation detection based on prior knowledge

作　　者：姜朝曦 周峻明 贝钰垠 

机构地区：[1]中国海诚工程科技股份有限公司,上海200031

出　　处：《自动化与仪器仪表》2015年第6期158-161 164,164,共5页Automation & Instrumentation

摘　　要：基于先验知识原理,本文提出一种全新的车道图像识别处理方式,分为如下几个部分,车道图像的预处理,基于先验知识的动态阈值二值化,基于Kalman滤波预测的Canny算子边缘检测,带直线拟合的Hough变换,自适应方式选取Hough变换的阈值。本文采用自适应方法获取Hough变换的阈值,并加入了直线拟合算法,准确的定位出直线位置,能够对有一定宽度和曲度的车道线进行更好的识别。同时利用上次定位的结果来限定当前车道的车道线搜索的图像区域,这样可以根据运动的动态信息缩小敏感区的范围,既可以可靠地定位车道线,又能够提高准确性,实时性。Based on priori knowledge, this paper gives a new image recognition approach which can be divided into below components. Pre-processing of lane image, automatic threshold based on priori knowledge, edge detection using Canny operator based on Kalman filter prediction, Hough transform with straight line fitting, the adaptive selection of Hough transform threshold. This paper adopts self-adaptive approach to obtain the threshold of Hough transform and joins a linear fitting method in order to locate the accurate line location, this method can be used to identify the lane with a certain width and curvature. Meanwhile the former result can be used to limit the latter searching area which could help to locate the lane accurately as well as enhance the efficiency.

关 键 词：先验知识 HOUGH变换 车道偏离 KALMAN预测 CANNY算子 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]