11:00 〜 13:00
[MGI31-P02] Structure-from-Motion Photogrammetry as a Low-Cost 3D Pointcloud Solution for Law Enforcement
キーワード：Geoforensic, SfM-MVS, photogrammetry
Reconstructing Accident and Crime Scenes in 3D, including environment details in 3D have been increasingly used as a complement and replacement of scenes' photographic evidences. This method is often carried out using Terrestrial Laser Scanners (TLS) which, combined with a camera, can create 3D scenes' reconstructions from which space measurements can be made and which can also be used to run simulations of flow dynamics, vehicle motion, human-being motion. Furthermore, it can even be used as training situations if fed through 3D virtual reality headsets.
Those techniques, although promising, are however limited by the availability of the equipment, with TLS for outdoor application starting with a price-tag >40k US$.
In the present contribution, the authors are therefore presenting a method applied to Earth Surface Processes to replace expensive TLS to an equipment set < 1000 US$: Structure-from-Motion Photogrammetry (SfM-MVS).
SfM-MVS relies on a set of photographs that are taken in spatial order, with overlap exceeding 60% in horizontal direction and at least 30% for two overlapping rows of photographs. The photographs, then taken, are fed into a sets of algorithms that calculate the position of the camera as well as the 3D scene, semi-simultaneously by updating one another using a statistical approach reducing the error. During the scene recording with the camera, it is essential to lay down sets of rulers and objects of known size in order to calibrate the 3D relation of the world to the reconstructed scene, and in order to estimate the error.
The 3D point-cloud can then be fed into open-source 3D point-cloud data handling software, like the open-source cloud-compare where various measurements can be made. The present work proposes a set of indoor and outdoor examples, which hopefully will inspire law-enforcement to use this method further and also build databases of cases that can be access online or via a computer for notably reconstruction and education purposes.
Those techniques, although promising, are however limited by the availability of the equipment, with TLS for outdoor application starting with a price-tag >40k US$.
In the present contribution, the authors are therefore presenting a method applied to Earth Surface Processes to replace expensive TLS to an equipment set < 1000 US$: Structure-from-Motion Photogrammetry (SfM-MVS).
SfM-MVS relies on a set of photographs that are taken in spatial order, with overlap exceeding 60% in horizontal direction and at least 30% for two overlapping rows of photographs. The photographs, then taken, are fed into a sets of algorithms that calculate the position of the camera as well as the 3D scene, semi-simultaneously by updating one another using a statistical approach reducing the error. During the scene recording with the camera, it is essential to lay down sets of rulers and objects of known size in order to calibrate the 3D relation of the world to the reconstructed scene, and in order to estimate the error.
The 3D point-cloud can then be fed into open-source 3D point-cloud data handling software, like the open-source cloud-compare where various measurements can be made. The present work proposes a set of indoor and outdoor examples, which hopefully will inspire law-enforcement to use this method further and also build databases of cases that can be access online or via a computer for notably reconstruction and education purposes.