Thanks to recent advances in deep learning (DL) and the increasing availability of large labeled/annotated datasets and trained network models, there has been impressive progress in the automated analysis of images from different scientific domains such as medicine, microbiology, astronomy and remote sensing. The automated analysis of archaeo-geophysical data is also considered important due to the large spatial extent of areas covered by landscape surveys using multi-sensor arrays driven by motorized carts and subsequently the large volume of collected data. In this work, a convolutional neural network (CNN) is built by Python 3.6 programming language using the Deep Learning Library of Keras with Tensorflow backends, a library that implements the building blocks for CNN. The network is trained from scratch adopting U-Net architecture to accomplish an automatic analysis of the archaeo-geophysical features with emphasis on ground-penetrating radar (GPR) anomalies. K E Y W O R D S archaeo-geophysics, convolutional neural networks (CNNs), deep learning, feature extraction, GPR (ground-penetrating radar), U-Net

Today, the advances in airborne LIDAR technology provide high-resolution datasets that allow specialists to detect archaeological features hidden under wooded areas more efficiently. Still, the complexity and large scale of these datasets require automated analysis. In this respect, artificial intelligence (AI)-based analysis has recently created an alternative approach for interpreting remote sensing data. In this study, a convolutional neural network (CNN) is proposed to detect clearance cairns, which are visible in today's landscape and act as important markers of past agricultural activities. For this aim, the U-shape network architecture is adapted, trained from scratch with an original labelled dataset and tested in various field sites, focusing on southern Sweden. Although it is challenging to tune the hyperparameters and decide on the proper network architecture to obtain reliable prediction, long-running experimental tests with this model produced promising results, with training and validation metrics of 0.8406 Dice-coefficient, 0.7469 Val-dice coefficient, and 0.7350 IuO and 0.6034 Val-IoU values, once trained with the best parameters. Thus, the proposed CNN model in this study made data interpretation quicker and guided scholars to focus on the location of the target objects, opening a new frontier for future landscape analysis and archaeological research.

No abstract

Highlights:• Complexity of large-scale Airborne LIDAR data: its processing, and interpretation emerges the necessity of automated analysis with novel techniques.• Detection and documentation of archaeological ruins, hidden in the forests of the Swedish landscape.