应用匹配滤波器的车道线恒虚警率识别方法  

Lane detection method of constant false alarm rate by using matched filter

作　　者：宋晓琳[1] 张三林[1] 张伟伟[1] 

机构地区：[1]湖南大学汽车车身先进设计制造国家重点实验室,长沙410082

出　　处：《电子测量与仪器学报》2015年第3期328-336,共9页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(51175159);湖南省科技计划(2013WK3024);湖南省研究生科研创新(CX2013B146)项目

摘　　要：针对多种道路场景下车道线的识别问题,提出了一种应用匹配滤波器和恒虚警率检测器的处理方法。根据车道线横截面的灰度轮廓特征,对图像局部区域,采用不同尺度的匹配滤波算子,实现不同环境条件下车道线区域的特征增强;利用可产生自适应阈值的恒虚警率检测器,提取车道线特征点,并通过随机抽样一致性方法拟合产生车道边界线。在车道线跟踪阶段,建立车道线参数数据库,分析当前帧车道线与数据库的匹配置信度,并动态更新该数据库。在嵌入式系统中,经过多种工况的试验验证,该算法能稳定地识别车道线,准确提取相关参数,处理速率达到30FPS。A processing method using the matched filter and constant false alarm rate detector is proposed to detect lanes under different road scenarios.According to the gray level profile of the lane cross-section, the matched filter operators with different scales are used to enhance lane markings in the local region of the images under different environments.Moreover, a constant false alarm rate detector is designed to generate an adaptive threshold for the extraction of lane markings.At last, the lane boundary fitting is accurately done based on Random Sample Consen-sus.During the lane tracking phase, the database of lane parameters is established, and current lane parameters are matched with those in the database to determine the confidence .Meanwhile the database gets update dynamical-ly.In the embedded system, the experiments under various conditions show that the lane detection algorithm can stably detect lanes and accurately estimate lane parameters with the processing frame rate of 30fps.

关 键 词：车道线识别 匹配滤波器 恒虚警率检测器 车道线跟踪 

分 类 号：TN713[电子电信—电路与系统] TP391[自动化与计算机技术—计算机应用技术]