摘要
针对高分辨率航空影像的特点,提出了基于角点检测的建筑物轮廓矢量化方法:首先,对航空影像进行多尺度面积形态学分割,获取建筑物的二值图像;其次,利用边缘跟踪技术,记录边缘点,在边界曲线上利用高斯函数计算其曲率,提取候选角点集,并由自适应支持区域确定的角点角度和一个动态曲率阈值代替固定的阈值筛选出正确角点;最后,以直角作为约束条件,对直线段进行规划和修整,得到最终的矢量化结果。实验结果证明,该方法具有运算速度快,精度高的优点,适合于高分辨率航空影像中建筑物的轮廓矢量化。
In this paper,a novel vectorization algorithm of building＇ contours is proposed：firstly,the multi-scale area morphological segmentation is carried out on the aerial imagery and the binary image is generated;secondly,the edge points are recorded by applying the edge tracking technology,according to the curvature calculated at a fix low scale of Gaussian filter for each point of the contour,the corner candidates are obtained;then,angles of corner candidates checked in an adaptive region of support and a dynamic curvature threshold is used instead of the fixed threshold to get the true corners;finally,a right angle constraint is introduced to modify the lines and vectorization is completed.With all procedures above,experiments show that the proposed algorithm with high accuracy and efficiency offers a suitable and robust solution to the contour vectorization of building in aerial imagery with high resolution.
出处
《遥感信息》
CSCD
2011年第5期95-99,共5页
Remote Sensing Information
关键词
高分辨率航空影像
多尺度面积形态学分割
边缘跟踪
角点提取
矢量化
aerial imagery of high resolution
multi-scale area morphological segmentation
edge tracking
corner extraction
vectorization
