{"id":73059,"date":"2021-11-05T09:38:45","date_gmt":"2021-11-05T08:38:45","guid":{"rendered":"https:\/\/icac.cat\/?p=73059"},"modified":"2021-11-05T11:46:31","modified_gmt":"2021-11-05T10:46:31","slug":"new-machine-learning-algorithm-detects-near-10k-archaeological-tumuli-in-galicia","status":"publish","type":"post","link":"https:\/\/icac.cat\/en\/actualitat\/noticies\/2021\/new-machine-learning-algorithm-detects-near-10k-archaeological-tumuli-in-galicia\/","title":{"rendered":"New Machine Learning Algorithm Detects Near 10k Archaeological Tumuli in Galicia"},"content":{"rendered":"

Archaeological tumuli <\/strong>are one of the most common types of archaeological sites and can be found across the globe. This is perhaps why many studies have attempted to develop methods for their automated detection.<\/p>\n

Their characteristic tumular shape<\/strong> has been the primary feature for their identification on the field and in LiDAR-based topographic data<\/a>, which usually takes the form of Digital Terrain Models (DTMs).<\/p>\n

The simple shape of mounds or tumuli is ideal for their detection using deep learning<\/a> approaches. Deep learning detectors usually require large quantities of training data (in the order of thousands of examples) to be able to produce significant results.<\/p>\n

However, the homogenously semi-hemispherical shape of tumuli<\/strong>, allows the training of usable detectors with a much lower quantity of training data, reducing considerably the effort required to obtain it and the significant computational resources necessary to train a convolutional neural network (CNN) detector.<\/p>\n

\"\"
Topographic data based on LiDAR. Image: Miguel Carrero-Pazos.<\/figcaption><\/figure>\n

This type of feature, however, presents an important drawback. Their common, simple, and regular shape is similar to many other non-archaeological features, and therefore studies implementing methods for mound detection in LiDAR-derived DTMs and other high-resolution datasets are characterised by a very large presence of false positives<\/strong> (objects incorrectly identified as mounds).<\/p>\n

\n
\n

The new algorithm provides a way forward for the detection of tumuli avoiding the inclusion of most false positives.<\/p>\n<\/div>\n<\/div>\n

GIAP<\/a> researchers Iban Berganzo<\/a> and H\u00e8ctor A. Orengo<\/a> faced this situation when carrying out the first stage of a study on the automatic detection of burial mounds in Galicia. A blog post<\/a> was recently posted on this initial investigation, in which almost 9,000 burial mounds were located. However, they were not all real burial mounds, as the results of the automated detection also included false positives.<\/p>\n

\"\"
Detected tumuli <\/em>in Galicia (Spain): (a) point distribution; (b) heat map. Author: Iban Berganzo (ICAC).<\/figcaption><\/figure>\n

After initial data validation was performed in collaboration with our colleagues Dr. Miguel Carrero<\/a> (University College London<\/a> \/ University of Santiago de Compostela<\/a>, GEPN-AAT<\/a>), Dr. Jo\u00e3o Fonte<\/a>\u00a0(University\u00a0 of Exeter<\/a>), and Dr. Benito Vilas<\/a>\u00a0(University of Vigo<\/a>), the team realised that from the ca. 9000 detected objects only ca. 7600 corresponded to real archaeological mounds. Although this was an excellent result<\/strong>, well below the percentage of false positives presented by similar studies, researchers thought they could improve the detection rate while decreasing the number of false positives.<\/p>\n

\n

During the summer, GIAP members Iban Berganzo<\/a> and H\u00e8ctor A. Orengo<\/a>, in collaboration with Dr. Felipe Lumbreras<\/a> from the Computer Vision Center (CVC)<\/a>, developed a new approach to reduce the number of false positives while increasing the detection rate.<\/p>\n<\/div>\n

After analysing the nature of the detected false positives, they developed a hybrid approach that mixes classical machine learning and deep learning<\/strong>. The objective was to obtain a more precise definition of archaeological tumuli in which not just the shape but also the multispectral characteristics of the objects will be considered when looking for tumuli.<\/p>\n

New results are now published as an open access paper<\/a> in the journal Remote Sensing<\/em><\/strong><\/a>, one of the top journals in the discipline. In this paper, researchers expand on the analysed data and information on this novel computer-based automatic detection initiative.<\/p>\n

\"\"
Figure published in the open access paper in Remote Sensing<\/em>, Berganzo et al., 2021: Graphical Abstract.<\/figcaption><\/figure>\n

The results that this new approach has produced are nothing less than spectacular:<\/p>\n