Photogrammetry and structure from motion
Photogrammetry is a technique that uses light rays captured by a camera. The technique fundamentally utilizes two two-dimensional (2D) images of the same object taken from different locations to create a three-dimensional representation of this object. Conventional photogrammetry requires the 3D location of control points or camera positions (locations and orientations). A stereo pair of images is processed to identify the corresponding control points in each image to calculate the 3D coordinates of the common points among the images as well as the camera locations and orientations by complex bundle adjustment algorithms.
Structure from motion (SfM) is an advanced photogrammetric technique that adopts the same basic principles of conventional photogrammetry, but SfM can determine the camera positions and orientations without specifying the 3D coordinates of control points. With SfM, the camera position is calculated using highly redundant, interactive bundle adjustment algorithms to extract the features from multiple overlapping images and to generate sparse point clouds. The sparse point clouds are then further processed to generate denser point clouds using the multiple view stereo (MVS) approach.
Application in digitizing rock slope and joint mapping
Any camera can be used for photogrammetry, such as digital single-lens reflex (DSLR) cameras, mirrorless cameras or cellphone cameras. However, better quality cameras and lenses can deliver better images, which yields better 3D model quality. Unmanned aerial vehicles (UAVs) or drones are becoming popular as a useful tool to acquire geometric data for rock faces.
A typical workflow when using photogrammetry to digitize rock slopes includes:
- placing the ground control points (GCPs) on or near the rock slope area.
- surveying the coordinates of the GCPs or camera locations.
- capturing multiple photos of the rock face.
- post-processing collected images in photogrammetry software.
Georeferencing, camera settings and image capture techniques affect the quality and accuracy of obtained models. The accuracy can be less than a few centimeters for 3D models with good georeferencing and practice.
Structural orientations, joint spacing and joint roughness can be measured on the 3D models of a rock slope for joint analysis. Figure 1 shows an example of a digital terrain model of rock slope with joint mapping. The model was constructed using images taken by an iPhone 12.