N-view reconstruction: a new method for morphological modelling and deformation measurement in volcanology

We present a method of reconstructing volcanic morphology using multiple digital views (N-view), captured at different angles around an object. This approach uses recent advances in the field of Computer Vision to provide accurate 3-D measurements of volcanic surfaces. Videogrammetry (digital image reconstruction) is used, as it is best adapted to numerical processing. The method is tested and now used in the laboratory on analogue volcanic cones. The method begins with calibrating the camera and finding image positions, using an accurate N-view calibration method. This is done by estimating internal and external parameters of the camera using several views of a specially constructed calibration target. The N-view reconstruction of the real object is then done by iteratively deforming an initial theoretical model of the surface. Laboratory tests show that reconstruction accuracy is about 10−4 m for a distance between the object and the camera of 0.5 m, and is potentially several orders of magnitude higher for surfaces of finer texture and using higher precision sensors. This is easily high enough to be useful for the accuracy required for morphological studies. It is also sufficient for monitoring most types of volcano deformation. The technique has the potential to detect morphology changes of the order of mm. Use of the method in the field requires a slightly different approach from that in the laboratory: textures and lighting are more variable, and the sensor and ground control point location and model calibration must be done differently. We provide case studies from laboratory tests and qualitative image analysis for two field cases: Piton de la Fournaise (Indian Ocean) and Santaguito (Guatemala). These illustrate the technique’s potential and explore problems of field application. Using current sensors, the method has the potential to provide sufficient precision for fine (mm–cm) scale reconstruction, and will represent a valuable, simple and flexible tool when compared with classical stereophotogrammetry techniques.