Hybrid MSRM-Based Deep Learning and Multitemporal Sentinel 2-Based Machine Learning Algorithm Detects Near 10k Archaeological Tumuli in North-Western Iberia

Iban Berganzo-Besga,Felipe Lumbreras,Miguel Carrero-Pazos,Benito Vilas-Estévez,Hector A Orengo,João Fonte

doi:10.3390/rs13204181

Abstract

This paper presents an algorithm for large-scale automatic detection of burial mounds, one of the most common types of archaeological sites globally, using LiDAR and multispectral satellite data. Although previous attempts were able to detect a good proportion of the known mounds in a given area, they still presented high numbers of false positives and low precision values. Our proposed approach combines random forest for soil classification using multitemporal multispectral Sentinel-2 data and a deep learning model using YOLOv3 on LiDAR data previously pre-processed using a multi–scale relief model. The resulting algorithm significantly improves previous attempts with a detection rate of 89.5%, an average precision of 66.75%, a recall value of 0.64 and a precision of 0.97, which allowed, with a small set of training data, the detection of 10,527 burial mounds over an area of near 30,000 km2, the largest in which such an approach has ever been applied. The open code and platforms employed to develop the algorithm allow this method to be applied anywhere LiDAR data or high-resolution digital terrain models are available.

Highlights

During the last 5 years, the use of artificial intelligence (AI) for the detection of archaeological sites and features has increased exponentially [1]
In conjunction with the information provided by the presence of false negatives (FNs) (35.58% of the test data), our results suggest that the approximate number of tumular features that could correspond to archaeological tumuli in Galicia approximates 14,626 (9422 estimated true positives (TPs) plus the estimation of those not detected according to the percentage of FNs)
The most recent approaches to the detection of archaeological mounds using LiDAR-derived data are usually able to detect a high percentage of the test dataset’s true mounds, but they include a large proportion of false positives (FPs)

Summary

Introduction

During the last 5 years, the use of artificial intelligence (AI) for the detection of archaeological sites and features has increased exponentially [1]. There has been considerable diversity of approaches, which respond to the specific object of study and the sources available for its detection. Deep learning (DL) algorithms, have been increasingly popular during the last few years, and they comprise the bulk of archaeological applications to archaeological site detection. DL approaches are diverse and include the extraction of site locations from historical maps [6] and automated archaeological survey [7], a high proportion of their application has been directed towards the detection of archaeological mounds and other topographic features in LiDAR datasets (e.g., [1,8,9,10,11]).

Objectives

Methods

Results

Discussion

Conclusion