Niloufar Rabbani
Interactive technology for construction site management.
Rel. Anna Osello. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Edile, 2026
Abstract
Interactive technology for construction site management Abstract This chapter presents the design, development, and current implementation status of a low-cost pipeline for automated construction progress monitoring that links fixed-camera site photography with a Building Information Model (BIM). The work is grounded in a case study of precast reinforced-concrete column erection at a construction site in Turin, Italy, monitored through a wide-angle camera that captures a new frame every five seconds. A structural-grid-based identification scheme is reused across the site drawings, the Revit BIM model, and the monitoring tool to give every column a single, unambiguous identity. Because the camera is not rigidly fixed and periodically shifts between a small number of discrete viewpoints, a drift-tolerant, per-date linking mechanism was developed to associate photographed columns with their BIM counterparts without requiring survey-grade camera calibration.
Planned and actual erection dates are compared automatically, and the results are fed back into the BIM through a pair of Dynamo graphs that write build dates onto column parameters and colour-code columns by schedule status. The chapter also describes an automated visual-detection layer that removes the manual marking step. Rather than training a detector — which would have required annotating a purpose-built dataset — a zero-shot, open-vocabulary object detector (Grounding DINO) was run locally on the site photographs using the text prompt “concrete column”, so that no model training, no annotated dataset, and no upload of the case-study site's photographs to any cloud service were required. Detections were matched to the previously linked column positions by testing, for each column, whether any pixel coordinate within a small square around its linked point falls inside a detection rectangle in the photograph of the same date. This approach identified 36 of the 41 columns (88%) automatically on their observed erection dates; the remaining five were completed by manual marking, giving a complete record for all 41 columns. The resulting dates were written back into the Revit model and used to colour-code the columns by schedule status. The chapter situates this work relative to the classical fixed-camera progress-monitoring literature (Lukins & Trucco, 2007; Ibrahim et al., 2009; Zhang et al., 2009) and discusses its contribution, limitations, and directions for further validation.
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Informazioni aggiuntive
Corso di laurea
Classe di laurea
Aziende collaboratrici
URI
![]() |
Modifica (riservato agli operatori) |
