车载视觉系统中机场场面标记牌识别算法

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。

Algorithm for airport surface tag brand identification in vehicle vision system

To achieve moving vehicle accurate position, this paper presents a taxiway tag brand identification method. First of all, preprocess the image and get the image binarization. On the basis of these processes, using mathematical morphology to mark the connected domain,the largest area is choosed as the candidate region . Then a new test method designed based on a single character area ratio of whole candidate region and aspect ratio of single character in the candidate region, can separate characters from the background successfully. Finally, use connected domain method to segment single character, use affine invariant moments to extract feature parameters, and choose the support vector machine (SVM) as the identifier. The experimental result shows that the character recognition rate achieves 96%, and the algorithm has strong adaptability for different images. It is the base of further moving targets positioning.