UAS-SfM Photogrammetry
I. Introduction
Structure from motion (SFM) photogrammetry is a revolutionary and low-cost technique for obtaining high resolution datasets to perform 3D aerial mapping with unmanned aircraft system (UAS). Pix4D SfM photogrammetry software works with all cameras and imagery data from UAS called the Sensefly eBee. The purpose of this report is to implement UAS, photogrammetry, Pix4D and GIS field technologies to perform the output Multi-View Stereo (MVS) workflow of 3D densified point cloud with RGB camera value, digital surface model (DSM) / digital elevation Models(DEM) and orthorectified image mosaic
II. Method
Automated aerial and close-range digital photogrammetry has become a powerful tool and popular used for three dimensional topographic modeling. The quality terrain data is extracted from overlapping stereo-pairs and enhanced. Digital photogrammetry also apply to number of geological problems as rock slope analysis, soil erostion, laboratory scale landscape models, etc. SfM has the same basic with stereoscopic photogrammetry resolving 3D structure from a series of overlapping offset images. The different is SfM uses a highly redundant and bundle adjustment procedure. SfM also can automatically generate sparse 3D point clouds from photosets.
Structure from Motion workflow
2.1. Image acquisition and key-point extraction
SfM can address the 3D location of matching features in multi images from different angles. The processing steps are the identification of features in images by using Scale Invariant Feature Transform (SIFT) method. Keypoints are identified over all scales with detail descriptions matched with large datasets. The number of keypoints is depended on texture and resolution of each image. If the spatial resolution increase and the distance between camera and features decrease, the spatial density and resolution of cloud points will enhance. SfM can obtain many images but three photos is minimum requirement. Small scale site with steep slope angle is the best choice to exclusive ground-based.
2.2. 3D scene reconstruction
After identifying the keypoints, the spare bundle adjustment system Bundler is use to estimate camera pose and extract a low density or sparse point cloud. Finally, triangulation is used to estimate the 3D point positions and reconstruct scene geometry to fix into a relative coordinate system. The bundler adjustment package can enhance density point cloud by implementing the Clustering View for MVS.
2.3. Post processing and digital elevation model generation
Transformation process from a relative to absolute coordinate system is achieved through ground control points (GCPs). A grid body transformation decomposed a rotation matrix and a scale factor. SfM generates 3D point clouds upward 10^3 point per m^2. The raw point clouds are directly interpolate into terrain model