Developments in the field of artificial intelligence have made great strides in the field of automatic semantic segmentation, both in the 2D (image) and 3D spaces. Within the context of 3D recording technology it has also seen application in several areas, most notably in creating semantically rich point clouds which is usually performed manually. In this paper, we propose the introduction of deep learning-based semantic image segmentation into the photogrammetric 3D reconstruction and classification workflow. The main objective is to be able to introduce semantic classification at the beginning of the classical photogrammetric workflow in order to automatically create classified dense point clouds by the end of the said workflow. In this regard, automatic image masking depending on pre-determined classes were performed using a previously trained neural network. The image masks were then employed during dense image matching in order to constraint the process into the respective classes, thus automatically creating semantically classified point clouds as the final output. Results show that the developed method is promising, with automation of the whole process feasible from input (images) to output (labelled point clouds). Quantitative assessment gave good results for specific classes e.g., building facades and windows, with IoU scores of 0.79 and 0.77 respectively.

This paper reports the knowledge process and the analyses performed to assess the seismic behavior of a heritage masonry building. The case study is a three-story masonry building that was the house of the Renaissance architect and painter Giorgio Vasari (the Vasari’s House museum). An interdisciplinary approach was adopted, following the Italian “Guidelines for the assessment and mitigation of the seismic risk of the cultural heritage”. This document proposes a methodology of investigation and analysis based on three evaluation levels (EL1, analysis at territorial level; EL2, local analysis and EL3, global analysis), according to an increasing level of knowledge on the building. A comprehensive knowledge process, composed by a 3D survey by Terrestrial Laser Scanning (TLS) and experimental in situ tests, allowed us to identify the basic structural geometry and to assess the value of mechanical parameters subsequently needed to perform a reliable structural assessment. The museum represents a typology of masonry building extremely diffused in the Italian territory, and the assessment of its seismic behavior was performed by investigating its global behavior through the EL1 and the EL3 analyses.

The increasing frequency and intensity of extreme weather events are accelerating the mechanisms of surface degradation of heritage buildings, and it is therefore appropriate to find automatic techniques to reduce the time and cost of monitoring and to support their planned conservation. A fully automated approach is presented here for the segmentation and classification of the architectural elements that make up one of the façades of Palazzo Pitti. The aim of this analysis is to provide tools for a more detailed assessment of the risk of detachment of parts of the pietraforte sandstone elements. Machine learning techniques were applied for the segmentation and classification of information from a DEM obtained via a photogrammetric drone survey. An unsupervised geometry-based classification of the segmented objects was performed using K-means for identifying the most vulnerable elements according to their shapes. The results were validated through comparing them with those obtained via manual segmentation and classification, as well as with studies carried out by experts in the field. The initial results, which can be integrated with non-geometric information, show the usefulness of drone surveys in the context of automatic monitoring of heritage buildings.

In this article, we present a new dataset designed for the semantic segmentation of images and point clouds of historic buildings, aimed at automating and accelerating the 3D modeling process in the scan-to-BIM context. The dataset includes five historic buildings and provides two types of data: images obtained from photogrammetric surveys and the corresponding georeferenced point clouds. Both datasets are accompanied by ground truth, identifying 10 classes representative of the main construction elements. Additionally, the dataset includes the intrinsic and extrinsic camera parameters and the transformation matrix to align the point clouds with the camera reference system. The annotation of the point clouds was performed manually, while the image annotation was generated through a projection process based on the labels assigned to the point clouds.

Despite GNSS (Global Navigation Satellite System) enables positioning, navigation and timing (PNT) almost everywhere, the development of applications like self-driving vehicles and indoor navigation requires extending accurate positioning to scenarios where GNSS either is not reliable or does garantee a sufficiently precise solution. Integrating inforamtion provided by different sensors is commonly accepted to be a quite viable way for such extension. In particular this work is part of a project aiming at investigating the positioning performance that can be obtained by integrating vision with radio-based systems and inertial sensors, which are commonly installed on many smart devices, such as smartphones. Furthermore, this work considers positioning in a collaborative scenario, where different interconnected platforms, i.e. unmanned aerial vehicles and pedestrians provided with smartphones, are moving on the same area. The results obtained in the considered tests show a good potential (submetric 2D positioning error) for what concerns the implemented strategies, where the integration of different technologies can ensure decent performance in a wider range of working cases.