The unmanned aerial vehicle (UAV) photogrammetric survey of an archaeological site has proved itself to be particularly efficient. In order to obtain highly accurate and reliable results, it is necessary to design carefully the flight plan and the geo-referencing, while also evaluating the indicators of the accuracy rate. Using as a test case a UAV photogrammetric survey conducted on the archaeological site of the Roman Amphitheatre of Avella (Italy), in this paper, we propose a pipeline to assess the accuracy of the results according to some quality indicators. The flight configuration and the georeferencing chosen is then be checked via the residuals on the ground control points (GCPs), evenly distributed on the edges and over the entire area. With the aim of appraising the accuracy of the final model, we will suggest a method for the outlier detection, taking into account the statistical distribution (both global and of portion of the study object) of the reprojection errors. A filter to reduce the noise within the model will then be implemented through the detection of the angle formed by homologous rays, in order to reach a compromise between the number of the usable points and the reduction of the noise linked to the definition of the 3D model.

The paper presents an innovative approach to support survey methods by applying AI algorithms to improve the accuracy of point clouds generated from UAV imagery. The work analyses different photogrammetric accuracy parameters in a first step, such as reprojection error and the intersection angle between homologous rays, verifying that a single parameter is enough to evaluate the accuracy of the photogrammetric restitution. Therefore, some of the calculated parameters were analysed through a Self-Organizing Map (SOM) to reach a compromise between the value of the variables analysed and the noise reduction associated with the 3D model definition. In the case study, it has been observed that the parameter that most influences the noise in the photogrammetric point clouds is the intersection angle.

Many studies on the semantic segmentation of cracks using the machine learning (ML) technique can be found in the relevant literature. To date, the results obtained are quite good, but often the accuracy of the trained model and the results obtained are evaluated using traditional metrics only, and in most cases, the goal is to detect only the occurrence of cracks. Particular attention should be paid to the thickness of the segmented crack since, in road pavement maintenance, the width of the crack is the main parameter and is the one that characterizes the severity levels. The aim of our study is to optimize the crack segmentation process through the implementation of a modified U-Net model-based algorithm. For this, the Crack500 dataset is used, and then the results are compared with those obtained from the U-Net algorithm, which is currently found to be the most accurate and performant in the literature. The results are promising and accurate, as the findings on the shape and width of the segmented cracks are very close to reality.

This work presents an experiment conducted on the settlement of Nuceriola, an archaeological site identified along the ancient via Appia, the road that starting from Rome passed through the city of Beneventum (Italy) in the direction of Brindisi. Different tools have been applied to enhance the detection of archaeological cropmarks. Since 2011, the Ancient Appia Landscapes Project has been working on the detection of the dynamics and the network of ancient settlement located along the Appia road in the east of Benevento, along with the cyclical elements and the human activities that influenced the evolution of landscapes. Starting from the photogrammetric results obtained from the processing of images from UAV, this document compares different types of mapping applying vegetative indices (VI) on raster data in order to point out archaeological evidence on land cultivated with wheat. For aerial image shots, a commercial quadcopter with RGB camera was used to identify the buried remains in archaeological settlements by means of visual recognition. The aim was to verify which types of mapping with VI can produce the best results for the display of archaeological finds, especially in terms of cropmarks. The case study shows that the use of only RGB cameras, without the addition of multispectral or thermal cameras, already allows the digital recording of buried archaeological remains through the application of appropriate filtering procedures of the colorimetric data and the vegetative indices.

No abstract