Archaeological Surface Ceramics Detection Using Low-Altitude Images Based on Weak Learners

Abstract

Over the past decade, remote sensing sensors and their products have been increasingly utilized for archaeological science and cultural heritage studies. In our study, we explored the application of several supervised machine learning classifiers using red-green-blue (RGB) and multispectral high-resolution drone imagery to evaluate their performance towards semi-automatic surface ceramic detection. The results indicate that using low-altitude remote sensing sensors and advanced image-processing techniques can be incredibly innovative in archaeological research. However, our study also revealed existing research limitations in detecting surface ceramics, which significantly impact the detection accuracy. Therefore, detecting surface ceramics using RGB or multi-spectral drone imagery should be reconsidered as an 'imbalanced data distribution' problem. A new and robust methodology needed to be developed to address this "accuracy paradox" of imbalanced data samples and optimise archaeological surface ceramic detection. Our study aimed to fill a gap in the literature by blending AI methodologies for non-uniformly distributed classes.