Artificial Intelligence for rapid mapping of potential archaeological features using Bag of Visual Words based image classifier

Abstract Integrating Artificial Intelligence technological advancements in archaeology has revolutionised automated feature detection, presenting a novel perspective on archaeological feature recognition and image interpretation. This approach reduces costs associated with ground data collection and enhances the reliability and productivity of large-scale archaeological mapping. Consequently, this study aims to explore feature detection and matching techniques in archaeological detection using Artificial Intelligence and Scale-Invariant Feature Transform and Oriented Fast and Rotated Brief algorithms, which are frequently employed in image processing applications as a feature descriptor within the Bag-of-Visual-Words framework. The high-resolution multispectral satellite SPOT image maps potentially hidden archaeological features in Bujang Valley, Kedah, Malaysia. The expected outcome involves presenting a BoVW model capable of accurately detecting hidden archaeological features within the generated maps, thereby providing valuable insights into the extent and distribution of archaeological remnants in the targeted regions.