Editing Shadow marks (section)

== AI applications ==
The collaboration of shadow mark analysis with [[artificial intelligence]] (AI) has paved the way for new opportunities in archaeological remote sensing.<ref name=":13" /> Shadow mark analysis had traditionally relied on the interpretation of aerial photographs to identify archaeological sites. Still, recent technological advances in machine learning and computer vision have led to automated shadow mark analysis with greater efficiency and accuracy. AI-based approaches can allow researchers to analyze countless datasets of aerial and satellite images to identify shadow marks with minimal human involvement.<ref name=":13" />

An evident and prominent development in this area is the deployment of [[Convolutional neural network|convolutional neural networks]] (CNN) in detecting and classifying shadow marks.<ref name=":14">{{Cite journal |last1=Argyrou |first1=Argyro |last2=Agapiou |first2=Athos |date=2022-11-26 |title=A Review of Artificial Intelligence and Remote Sensing for Archaeological Research |journal=Remote Sensing |language=en |volume=14 |issue=23 |pages=3–19 |doi=10.3390/rs14236000 |doi-access=free |bibcode=2022RemS...14.6000A |issn=2072-4292}}</ref> CNNs can distinguish between genuine archaeological shadow features and other artifacts caused by clouds, vegetation, or urban features.<ref name=":14" /> Moreover, researchers have recently demonstrated improved accuracy in the automated detection of shadow marks by training AI models with datasets containing known archaeological sites.<ref name=":13" /> Researchers have recently used unsupervised learning methods such as clustering algorithms to segment aerial imagery and extract shadow features of possible underground features.<ref name=":13" /><ref name=":14" />

Another significant development involves applying 3D photogrammetry and light simulation to improve the visibility of shadow marks. Virtual reconstructions of lighting conditions allow researchers to simulate and recreate both sun angles and the corresponding shadows, allowing them to visualize how an archaeological site would appear under different lighting circumstances. This technique has been beneficial in instances when real-world shadow marks cannot be detected due to seasonal or weather conditions.<ref name=":12" />

AI-driven remote sensing has also begun to employ multi-spectral data processing concurrently with shadow analysis.<ref name=":13" /> By applying spectral indices like the [[Normalized difference vegetation index|Normalized Difference Vegetation Index]] (NDVI), researchers can explore patterns in an object’s shadow mark, establishing a means to differentiate between shadows associated with burial remains from those generated by vegetation differences.<ref name=":13" /> The separation allows researchers to filter out false positive occasions, particularly in dense forest areas where the distinction between tree and archaeological shadows can be problematic.<ref name=":14" />

These AI methodologies for shadow mark detection are still limited by their reliance on high-quality training datasets that are not always available throughout the world.<ref name=":13" /> Another concern is that variable landscapes can also complicate and further diversify the types of shadows produced over time, meaning that AI processes of shadow recognition will require continuous development and improvement.<ref name=":14" /> Regardless, as more datasets become available, it is anticipated that AI methodologies will become a routine part of chronological remote sensing and shadow analysis for archaeological inquiry.<ref name=":13" /><ref name=":14" />