VALIDATING PHOTOGRAMMETRIC ORIENTATION STEPS BY THE USE OF RELEVANT THEORETICAL MODELS. IMPLEMENTATION IN THE "ARPENTEUR" FRAMEWORK

Abstract

The new advance in photogrammetry using the automatic procedures such as the famous algorithm which was proposed by David Lowe (Lowe, 2004) features descriptors and matching (SIFT) and then the recent development of external orientation (Nister (Stewenius et alii, 2006) or Snavely (Snavely et alii, 2010)) have changed drastically the way of measuring space with photogrammetry. The complexity of the process and the huge quantity of processed data (thousands of photographs) makes difficult validating the different process steps. We propose in this paper several theoretical model generation methods in order to validate the complete photogrammetric orientation process. A theoretical photogrammetric model generation has been developed in order to produce photographs, photo orientation, 3D points and 2D observations according to some defined camera and a parametric photograph distribution in the scene. In addition the use of synthesis image software generation as POV-Ray allow us to generate set of photographs with pre-computed internal and external orientation in order to check the whole pipeline from feature extraction to Photographs External Orientation. We apply this model generation approach to several typical geometry of photogrammetric scene, stereo, parallel triplet, parallel strip and convergent models.