Methodologies for assessing the quality of 3D models obtained using close-range photogrammetry

Abstract

Although reality-based models are widely used to describe the geometric surfaces of an entity in a digital space, a systematic and universally recognised treatment of issues such as accuracy is lacking. The topic is certainly complex as this analysis should involve not only shape approximation but also other attributes (e.g., colour). Wanting to limit ourselves to geometry alone, this work proposes solutions for assessing the quality of photogrammetric models, differentiating them according to possible scenarios: sometimes, homologous models obtained using different techniques and technologies are available. In these cases, a comparison between digital reconstructions can serve to effectively quantify accuracy; more often, no terms of comparison are available, and one is forced to derive indicators from the same photogrammetric process to describe quality. We propose for this scenario a statistical analysis on the covariance matrix of the estimated coordinates for the tie points. The main goal is to provide a range of possible approaches to the conscious management of survey data.