La Torre della Marina di Vietri sulla Costa d’Amalfi: test statistico-predittivi di dati fotogrammetrici

Abstract

[EN] The focus of the work is on close-range photogrammetry and mainly on the low-cost technologies, experimented in the survey of Tower of Marina di Vietri, a historical building erected in the sixteenth century at Vietri sul Mare in the Province of Salerno. The general objective is to codify a methodology for objectifying the comparisons of the results; hence, the research starts from an original analysis conducted on the returned orthophotos by several photogrammetric paradigms. To estimate its reliability and precision, we proceeded, at first, with the extension and application of an error propagation law and then with the validation of the comparison according to a predictive type test. The first results are presented here after a study on raster images generated according to different algorithms. Subsequently, on each graphic product, 73 points were identified and for each of the relative coordinates the deviation are evaluated by verifying them with the value of the standard deviation. Consequently, for the purpose of greater validation of the methodology, a predictive test was implemented with the aim to confirm the criterion used for the comparison and to guarantee, in probabilistic terms, the values analysed; finally, a further three-dimensional analysis was conducted directly on the 3D models. It is banal to observe that aleatory –subjectivity, etc.– which often characterizes already known approaches, in any case assumed as a starting point, does not always allow to obtain generally valid results and, therefore, extendable beyond the single case study. The implementation of this first test, otherwise, proved to be extremely valid in the survey for architecture, both for an absolute evaluation of the quality of the individual results and for the possibility to estimate, also predictively, the relative effectiveness of the method used. This guarantees the evaluation of the relative percentages of errors in probabilistic terms.