一种航空影像密集匹配点云的迭代中值滤波算法

An iterative median-value filtering algorithm for aerial-image dense matching point cloud

为有效解决基于航空影像密集匹配生成的点云中含有大量的冗余点和噪声信息,影响表面模型(DSM)的质量和后续数字正射影像(DOM)生成效果的问题,提出了一种航空影像密集匹配点云的迭代中值滤波算法。通过引入邻域法向量求和的思想定义空间圆柱体,并以此作为点云剔除的约束准则,判断空间圆柱体内是否需要获取中值点,并结合距离统计分析的方法剔除少量离群点,迭代直至相邻两次输出点云数量差异少于设定阈值。采用具有重叠区域的两类点云数据进行对比实验,结果表明:提出的方法能够有效剔除摄影测量密集匹配点云的噪声,对于存在厚度的冗余点云进行光顺平滑,能够保留完整的边缘特征。

The generated point cloud through dense matching based on aerial-image contains massive redundant points and noise information,which influences the quality of surface model(DSM)and the effect of digital orthophoto(DOM)generation.To solve this problem,an iterative median-value filter algorithm for aerial-image dense matching point cloud is proposed.The spatial cylinder is defined by the idea of neighborhood normal vector summation,and it is used as the constraint criterion of point cloud elimination to judge whether the median point is needed in the spatial cylinder.The presented algorithm removes a small number of outliers with distance statistical analysis method until the number of adjacent two output points is less than the set threshold.Comparative experiment using two classes of point cloud data with overlapping regions shows that the proposed method can effectively eliminate the noise of the photometric measurement dense matching point cloud,smooth the redundant point cloud with thickness,and preserve complete edge features.