一种基于对齐度的多约束车辆特征匹配方法  

Fast vehicle feature matching algorithm based on alignment metric

作　　者：吴振华[1] 吴晴[1] 郑榜贵[1] 

机构地区：[1]北京工业大学电子信息与控制工程学院,北京100124

出　　处：《电子测量技术》2009年第5期74-76,共3页Electronic Measurement Technology

基　　金：北京市教委科技创新平台项目支持(05002011200701)

摘　　要：针对基于双目立体视觉的汽车辅助驾驶系统,研究了一种基于对齐度的多约束快速匹配算法。对齐度准则利用图像对间的交互方差作为相似测度,不需要两幅图像的灰度线性相关性并且不受图像灰度差异的影响,比相关度准则有更强的适用性。确定左右两幅图像中的路面区和目标车辆后,采用Harris算子提取出目标车辆的特征点。匹配前利用角点的特征值约束去除不易匹配角点,然后采用对齐度准则进行匹配,最后根据视差梯度约束去除误匹配。实验结果表明了该方法的可靠性,能够获得较满意的检测结果。In the field of automotive driver assistance systems and binocular stereo vision, it studied a fast matching algorithm. During the process AM（Alignment Metric）is used as the matching criteria. AM is subject to the non-linear and other interferences of the images, and thus more reliable. During the whole process, the first aim is to position the target vehicle, then adopt Harris operator to extract feature points. After some not-easy-matching features are removed by eigenvalue ratio constraint, AM and disparity gradient constraint is used to complete the matching process. Experiment shows that this method is reliable and can achieve some satisfactory results.

关 键 词：双目视觉 特征点 匹配 对齐度 视差梯度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]