车载视觉系统中机场场面标记牌识别算法  

Algorithm for Airport Surface Signboard Identification in Vehicle Vision System

作　　者：李露[1] 王帮峰[1] 丁萌[1] 

机构地区：[1]南京航空航天大学民航学院,南京211100

出　　处：《科学技术与工程》2014年第17期234-239,共6页Science Technology and Engineering

基　　金：国家自然科学基金(61203170);中国民航科技项目(MHRD201124);中国博士后基金特别资助(2013T60539)资助

摘　　要：为实现移动车辆精确定位,提出了一种基于机场滑行道标记牌识别的方法。首先对图像进行预处理和二值化,利用数学形态学连通区域最大面积进行分割,实现候选区域(candidate region,CR)预确定;然后设计了一种基于CR区域单个字符连通域的外接矩形面积所占该标记牌牌面总面积的比值和字符宽高比的检验新方法,成功地将字符从背景中分离;最后采用连通域方法分割出单个字符。字符特征提取采用仿射不变矩的方法,选择支持向量机作为识别器,识别率达到96%。经验证,算法可较好地识别出标记牌字符,针对不同图像背景有较强的适应性,为下一步确定移动车辆的位置打下坚实基础。In order to achieve moving vehicle accurate positioning, an identification method of taxiway signboard was presented is presented. First, the images is preprocesed and the binarization is gotten. On the basis of these processes, a mathematical morphology was used to mark the connected domain, in which the largest area was chosen as the candidate region. Then a new test method was designed based on the area and aspect ratio of a single char- acter in the whole candidate region, which can separate characters from the background successfully. Finally, con- nected domain method is used to segment single character, used affine invariant moments to extract feature parame- ters, and chosen the support vector machine （SVM） as the identifier. The experimental result showed that the character recognition rate was 96%. The algorithm had strong adaptability for different images. This research is the basic of further moving targets positioning.

关 键 词：滑行道标记牌 候选区域 连通域分割 仿射不变矩 支持向量机 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]