基于多特征集成的卡口图像前车窗的定位算法  

A vehicle window localization algorithm based on multi-feature integration in traffic bayonet images

作　　者：曲宝珠 曹国[1] 刘宇[1] 周丽存 QU Bao-zhu;CAO Guo;LIU Yu;ZHOU Li-cun(School of Computer Science and Engineering,Nanjing University of Science and Technology,Nanjing 210094, China)

机构地区：[1]南京理工大学计算机科学与工程学院,南京210094

出　　处：《信息技术》2016年第12期19-24,29,共7页Information Technology

基　　金：863计划项目(2013AA014604)

摘　　要：随着高清卡口系统的普及,其所提供的静态卡口图像对公路交通安全建设有着重要的作用。针对该类图像,文中提出一种多特征集成的前车窗自动检测算法。首先,对图像进行车体检测,车牌定位,根据车牌与车窗的几何关系得到车窗粗定位的图像;然后用车窗直线特性和形状特性对车窗粗定位后的图像进行区域划分,将其分成几个候选区域;再对候选区域分别提取三个特征:形态学特征、形状特征、密封条光谱特征;最后建立多特征集成函数来判定车窗区域。实验表明,提出的方法具有较高的检测精度,同时当车型、车色以及光照变化较大时,该算法仍有较好的鲁棒性和适应性。With the popularity ol high-delinition bayonet system, the static bayonet images play an important role on the construction ol the highway traffic salety. To solve the problem ol vehicle Iront window localizationin traffic bayonet images, this paper propose an algorithm based on multi-leature integration. Firstly, it identilies the body ol vehicle and detect the license plate. According to thegeometric relationship between thelicense plate and the windscreen, the candidate region ol the lrontwindow can be lound. T h e n , the region is divided into severalareas based on the linear and shapecharacteristic ol the vehicle ’ s windscreen. Three features are extracted in these potential window areas :morphological leatures, shape leatures and seal spectral leatures. The linally, the vehicle window can belocated by integration ol these leatures. The experimental results demonstratethat the proposed algorithm has a high detection performance. Meanwhile, the method is robust against dillerent lighting conditionsand various vehicle types.

关 键 词：智能交通 车窗定位 形态学建筑物指数 多特征集成 卡口图像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]