Methods to accurately capture 3D space are time intensive and often require bespoke equipment. I've investigated a method of capturing the impression of a place rather than the precise outcomes only laser scanning can provide. The solution utilizes stabilization and photogrammetric technologies to produce models of spaces requiring minimal equipment and time. The following outlines the technology used and it's outcomes.
Photogrammetry is a technique of interpolating photographs to determine the location of points in 3d space. The underlying method at hand, known as Structure from Motion (hereby SfM), cross references a series of photographs that outline a sequence of transformations around or through the subject matter. In doing so, one can document the parallax shift between figures to predict depth.
The value of SfM is two fold. As a typical use, it provides measured information about the subject in terms of distances between many features from finite calibration images for sake of performing a conventional survey. The more extraordinary use case is the generation of interpolated images via 3D reconstructions. Photogrammetry is by no means new as it can be wholly performed manually. However, current technology is far more forgiving with input and provides high fidelity 3D geometry.
SfM would normally require perfectly consistent images for feature detection and comparison, often calling for the use of DLSR Cameras and manual shooting. Now, even compressed video can be subsampled and usable within the reconstruction of space.
One can see the articulation of the facade is well expressed, though the model is not to measure from, it no doubt captures the essence of the landmark.
Through use of the Osmo Mobile 2 (a consumer grade gimbal) with a modern smartphone, one can produce high fidelity stabilised footage of an object, space or place. Thanks to the software enabled feature detection and tracking (Via Agisoft Metashape for these examples), one can stay centred on a subject whilst moving around it. As a field test I've captured the corner of Sydney's Queen Victory Building using a snake like approach to best capture multiple angles in a single shot.
The results are very surprising in terms of fidelity given the poor lighting conditions for SfM and the use of compressed video. Due to the stabilisation provided via the gimbal, each frame can easily be matched with the previous as overlap is ensured and blur is almost entirely removed.
