Deep learning based automated analysis of archaeo‐geophysical images

Abstract

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

GPR Data Interpretation Approaches in Archaeological Prospection

“…Few studies exist in the literature exploring the automatic GPR data feature extraction capability of DL algorithms. An example utilizes a modification of CNNs, known as Fully Convolutional Networks (FCNs), that performs image segmentation through the U-net architecture [41]. In this study, the trained model takes as an input a GPR C-scan and outputs the linear features in segments that are attributed to structures.…”

Combined Detection and Segmentation of Archeological Structures from LiDAR Data Using a Deep Learning Approach

“…Using deep CNN for archaeological prospection of LiDAR derived-terrain (Caspari and Crespo 2019;Gallwey et al 2019;Küçükdemirci and Sarris 2020;Soroush et al 2020;Trier, Cowley and Waldeland 2018;Verschoof-van der Vaart et al 2020;Verschoof-van der Vaart and Lambers 2019) is in its infancy, and to our knowledge, these studies have not evaluated the object-segmentation abilities of the CNN, except the evaluation of Mask R-CNN for simple circular-based landforms (Kazimi, Thiemann & Sester 2019;Kazimi, Thiemann & Sester 2020). In the present study, we assess the contribution of deep CNN to the combined detection and segmentation of archeological structures for further (semi-)automatic characterization.…”

Learning from Ground Penetrating Radar to identify ancient buried structures

“…Deep learning applications in GPR data derived from archaeological prospection are currently unexplored. The recent study conducted by Küçükdemirci and Sarris in [96] is the closest and only one found towards the direction that this Ph.D. research is heading. Küçükdemirci and Sarris applied semantic segmentation using U-Net to identify buried structures in C-scans.…”

“…However, this is considered only the beginning, with many things left to improve, investigate, and try. Aside from classification tested here, image segmentation is a very promising direction for GPR images, as shown in the recent study by Küçükdemirci and Sarris [96]. A future direction could be the application of segmentation in 3D GPR volumes to extract 3D of the subsurface, a GPR representation that is currently lacking.…”

“…As it was described in subsection 4.5, a sufficient number of training data examples is required to describe the input space well enough, but practical guidelines on the smallest efficient number of images required are of lacking. For this reason, the size of the datasets used in similar GPR studies (like in [76] and [96]), as well as popular datasets used for training CNNs (e.g., MNIST [206] and CIFAR-10 [207]) were used as guides in filling a sufficient number of images. This has led to a minimum threshold of 5000 images per class as a starting point.…”