街景影像中的斑马线检测与定位  

Zebra-crossing detection and positioning in street view image

作　　者：闫利[1] 黄亮[1] 

机构地区：[1]武汉大学测绘学院,武汉430079

出　　处：《测绘科学》2015年第11期168-171,181,共5页Science of Surveying and Mapping

基　　金：中央高校基本科研业务费专项资金项目(2042014kf0084);精密工程与工业测量国家测绘地理信息局重点实验室开放基金项目(PF2013-11)

摘　　要：针对街景影像中斑马线的自动检测与定位识别率和定位精度不高的问题,文章引入俯视投影与霍夫直线检测来改进基于双极系数的斑马线识别算法,并设计球面全景单像测量方法对斑马线进行粗略定位:对全景图像上路面区域进行俯视投影,使图像各处比例尺保持一致;对路面俯视图分块计算以生成双极系数图,并用形态学中闭运算与开运算进行处理;对候选区域进行霍夫直线检测,利用直线长度与方向的统计信息来排除干扰;最后使用单像量测方法对斑马线进行自动地理定位。实验结果证明,该方法对街景中的斑马线具有较高的识别率与定位精度。Aiming at the problem that the precision of automatic detection and positioning of zebracrossing in panorama image is low,the paper introduced the vertical-view projection and Hough line detection to improve bipolarity-based zebra-crossing detection algorithm,and designed the single-panorama measuring method to get a rough position of the detected zebra-crossing：firstly,the region of road in latitude-longitude image was projected to a horizontal plane in order to get the same scale;next,bipolarity image,which is binary,was generated through using blocks of the vertical-view image followed by opening-and-closing operation of morphological processing;then,Hough line detection was carried out to the candidate area,and the statistical information of the linear length and direction was used to eliminate the interference region;finally,the automatic geo-positioning for the zebra-crossing was implemented with mono-image measuring algorithm.Experimental result showed that the method could have high precision of identification rate and positioning to the zebra-crossing in street views.

关 键 词：街景影像 斑马线检测 路面俯视图 双极系数 霍夫直线 粗略定位 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]