Accurate documentation of traffic accidents is a prerequisite for accident research as well as traffic jurisdiction. As an important part of accident documentation, stereo-photogrammetry is recognised to be an excellent tool for providing accurate and complete scaled maps of accident scenes. However, due to its relatively high expense, it is usually only applied in cases of severe accidents. In contrast, single-view photogrammetry which is based on photographs taken with “non-metric” cameras and on on-the-job calibration requires little installation at the accident scene and provides adequate accuracy, because camera calibration and plotting of scaled maps can be performed by making use of computer-assisted image analysis. As the method basically consists in a simple perspective rectification single-view photogrammetry as such is restricted to plane accident sites. In this paper the method is demonstrated and its accuracy discussed.
ACCURATE AND COMPLETE documentation of traffic accidents is the basis of accident research and analysis. Likewise, it is important for traffic jurisdiction.
The documentation should usually include a sketch of the accident site where the positions of the involved vehicles, associated skid marks, all relevant distances and other details of importance are documented. Ideally, such a sketch is a scaled map, typically 1:100, and is produced with the aid of stereophotogrammetric methods, which warrant the desired accuracy.
Accordingly, the city police of Zurich introduced stereophotogrammetry as a routine tool for accident documentation purposes as early as 1934 (1)*. Apart from its value for accident jurisdiction, stereophotogrammetry proved to be indispensible for a reliable accident reconstruction and analysis.
While the precision of stereophotogrammetry as applied today is unrivalled by other methods, a disadvantage exists in its relative expense such that in minor accidents where minimal legal consequences are expected its application is not justified. Yet, the analysis of such accidents is often particularly significant in accident research, e. g. when biomechanical tolerance limits are the subject of research.
In most cases, however, 35mm photographs are taken. The question therefore arises, to what degree quantitative information can be recovered from a single planar view. With the availability of inexpensive, small and easy-to-use image analysis hardware, the possibility of producing scaled accident sketches from a single photograph was investigated.
From a single planar view, of course, a true spatial reconstruction of a scene is not possible without having further information. In most cases, however, an accident site can be assumed to be, at least piecewise, planar with sufficient accuracy. Under these circumstances, the photogram-metric problem reduces to a simple perspective rectification because linearity of the mapping can be assumed for all practical purposes with present photographic equipment.
In this paper, the single-view photogrammatric method is presented, its accuracy and limitations discussed and a typical example is demonstrated.