【摘要】 为了缩小基于视频的数字城市三维重建中建筑物特征点的提取匹配的搜索范围,需要对目标建筑物区域进行快速提取。以矩形建筑物为例,首先对预处理后的图像做Hough变换,获取直线段集合;然后,在分析建筑物轮廓线的相互位置关系时,利用提出的点密度和Hough空间中计数器的数值,筛选出直线段集合中能表现建筑物轮廓信息的直线段。实验结果表明,运用点密度能去除干扰线段,算法对建筑物上下轮廓和左右轮廓采用不同的提取策略,在由于拍摄角度而造成的建筑物畸变的情况下,能快速准确地提取出矩形建筑物区域。更多还原

【Abstract】 In order to reduce the scope of characteristic dot extracting and matching in 3D reconstruction of digital city based on video,an auto building regional extraction algorithm was proposed.It attained lines array using Hough transform for the preprocessed image,and applied DotDensity proposed and the values of counter in Hough space analyzing relation between contours’ geometrical position to get the contour line segments.Experimental results present that using the proposed DotDensity can eliminate disturbing lines effectively and the approach can get acceptable results fast from images using different extracting methods for vertical and horizontal boundaries with less complicated backgrounds even if the building deforms in picture due to shooting angle.更多还原