DSM generation on the shade of tree area of aerial photogrammetry

Abstract

In aerial photogrammetry, a shooting area is occasionally covered by tree. The shade of trees prevent us from obtaining spatial information. The texture and color of the tree can be so homogenous that it is difficult to accurately extract the feature points. This problem makes it difficult to generate a DSM using aerial photogrammetry. This study attempted to determine a method for improving the accuracy of 3D information obtained from under a shade of tree. We used photos that were shot from under the tree, and combined scale-invariant feature transform (SIFT) and structure from motion (SfM) algorithms to conduct image-based modelling through the reconstruction of spatial geometry to produce a 3D point cloud. Data fusion was conducted to generate an entire DSM by integrating point cloud into a DSM map by using aerial photogrammetry. We assessed the accuracy of a DSM of the entire study area to extract feature points from study area as the control points in a data fusion of the point cloud and the DSM produced by aerial photogrammetry.