深度图像分割的城市区域倾斜影像密集匹配点云滤波算法

倾斜摄影测量作为一个新兴领域发展势头迅猛,在众多领域得到了广泛应用。但倾斜影像密集匹配点云处理技术研究却相对较少。倾斜影像密集匹配点云分布不均匀、表面粗糙,因而传统的激光扫描点云处理算法在用于倾斜影像密集匹配点云处理时的适用性较低。本文从倾斜影像密集匹配点云特点出发,提出了一种利用点云高程信息生成深度图像提取建筑物非连通区域,在全局范围选取种子点实现多种子点区域生长的点云快速滤波算法。实验结果表明,该算法滤波效果好、速度快,可以改善密集匹配点云部分地物底部边缘不清晰引起错分和区域生长无法分割建筑物非连通区域的问题。

Filtering Urban Point Cloud from Dense Image Matching Based on Depth Image Segmentation

Oblique photogrammetry is a new and rapidly developing technique which is widely used in many fields,but the study of dense matching point cloud processing technology is less.The typical point cloud filtering algorithms for laser scanning data are not suitable for dense matching point cloud because of the inhomogeneous point distribution and rough surface in dense matching data.In the light of the characteristics of dense point cloud,a fast point cloud filtering method is proposed.Firstly,depth image is generated from elevation data where the disconnected area is extracted using spatial features.Then,the depth image is segmented by region growing using multiple global seeds.Experimental results show that this algorithm is effective and efficient in filtering dense matching point cloud,especially in solving the problem that the bottom edges of the objects are not clear and the problem that region growing cannot perform well in the disconnected area in the buildings.