Alias not configured

[CLS] - Introduction

Cultural Landscapes Scanner (CLS)
Earth Observation and automated detection of subsoil undiscovered cultural heritage sites via AI approaches



Sub-surface, yet-undiscovered cultural heritage sites (such as buried ancient structures and monuments) can be identified on remote sensing data from a variety of sensors in the form of anomalies or traces detectable on bare soils, crops, and vegetation. The current availability of free remote sensing datasets through platforms, like Copernicus, is unprecedented. Such datasets have huge potential and are already amply used within the worldwide cultural heritage community thanks especially to the availability of time-series imagery. However, the extraordinary proliferation of data has posed significant hurdles in terms of managing, processing and interpreting them to the point that the quantity of data is not manageable by traditional ‘human’ visual interpretation.

The new challenge in cultural heritage remote sensing scholarship therefore is to develop or improve instruments that can facilitate automatic detection of objects of interest. This project takes charge of it by supporting the development of specific methods that look at automatically identify specific cultural heritage objects and patterns related to anthropogenic interference on landscapes in the past using the latest breakthroughs in Artificial Intelligence (AI). This project will also considerably expand existing approaches to the identification of ancient land division systems —and more generally of landscape patterning— by automating procedures of similarly-oriented linear feature detection.



Cultural Landscapes Scanner (CLS) is developing new methods based on Artificial Intelligence for processing different types of Remote Sensing data to semi-automatic identify cultural heritage sites. To achieve this research agenda this study focuses on the following issues:
- The barriers to the cross-fertilisation between Artificial Intelligence and cultural heritage, which are related to the quality and the access of suitable training datasets, often limited in span and not publicly available.
- The unavailability of a way to compare different object detection and semantic segmentation methods for the identification of cultural heritage sites. This issue is a consequence of the absence of a publicly available benchmark dataset on which testing the different Artificial Intelligence techniques.
- The lack of a standard set of measure to assess the performance of the different methods since each research group uses different performance evaluation approach based on their expertise and on their own unique dataset.



CLS project aims at setting a benchmark in the use of remote sensing data for the automatic identification of various classes of undiscovered cultural heritage sites through the integration of cutting-edge Artificial Intelligence techniques with archaeological research and fieldwork.
The foundation of its research agenda is embodied by the introduction of a first publicly available multi-modal dataset and by attempting for the first time to solve the problem of the non-stardisation of performance metrics. This benchmark dataset will contain Sentinel 2 multi-spectral images and LiDAR data of the archaeological landscape of Aquileia (Italy), a major city of the Roman Empire.
State-of-the art object detection and semantic segmentation methods will be explored to establish how the granularity of the detection affects the quality of the prediction from an archaeological point of view. To assess the performance of the different models, a set of different metrics will be introduced with the collaboration of landscape archaeologists in order to set a first reference standard for promoting objective cross-study measurements.
Archaeological ground-truthing will be undertaken in the case study area to verify the predicted results provided by the developed AI methods. The data collected during the ground-truthing activities will be used to improve the identification performance of the proposed methods.





Arianna Traviglia. PI

Project idea development and the heritage aspects related to the creation of the benchmark dataset.



Marco Fiorucci

Design of object detection and image segmentation methods for cultural heritage sites identification.



Paolo Soleni

Machine learning engineer: coding, models and experiments development



Marina Ljubenovic

Image processing for deblurring and denoising Sentinel-2 multi spectal data





The Cultural Landscapes Scanner pilot-project is the result of a partnership between IIT’s Centre for Cultural Heritage Technology and European Space Agency (ESA) as part of the ESA co-funded research programme Discovery & Preparation.