- Archaeological News
-
Image Credit: Bilaterian, CC BY-SA 4.0, via Wikimedia Commons
AI and Spectral Imaging Improve Identification of Ancient Ruins
A new study has introduced a high-precision method for identifying ancient human ruins by combining visible near-infrared spectroscopy with deep learning. The approach aims to distinguish archaeological remains from surrounding natural sediments, a task that is often difficult because many ruins gradually blend into their environment over time.
The research focused on ancient sites in Central China, a region with a long and continuous history of human settlement. Researchers collected more than 14,000 spectral samples from 27 archaeological sites, covering remains such as ash pits, burials, moats, rammed earth structures, architectural debris, loess, paleosols, and lacustrine sediments.
Using a deep learning model known as ResNet50, the team achieved a classification accuracy of 94.86%. When combined with Standard Normal Variate preprocessing, which reduces noise and scattering effects in spectral data, accuracy increased to 96.60%.
The study shows that artificial intelligence can support archaeological fieldwork by improving the speed and reliability of identifying buried or visually unclear cultural remains. This method may reduce reliance on manual judgment and provide a more systematic tool for archaeological surveys.
The authors note that the current dataset is mainly based on Central China and should be expanded to other regions, climates, and time periods. Even so, the results demonstrate the growing potential of combining spectral imaging, remote sensing, and AI in archaeological research.
Published on: 26-04-2026
Edited by: Abdulmnam Samakie
Source: npj Heritage Science
