Revolutionizing Archaeology: Unleashing AI's Power

Table of Contents

1. Introduction
2. Background and Training
3. The Power of AI in Archaeology
4. Detecting Archaeological Sites with AI
   1. Pattern Recognition vs. Deep Learning
   2. Examples of Successful Detections
   3. Challenges and Limitations
5. Scaling Up: From Local to National Coverage
   1. Lidar Availability and National Mapping Program
   2. Future Opportunities and Research Potential
6. Digitizing Historic Mapping with AI
   1. Data Gathering and Selection
   2. Mapping Orchards: Results and Insights
   3. Loss of Orchards and Implications for Land Management
   4. Next Steps and Research Opportunities
7. Conclusion

The Power of AI in Archaeology and Historic Mapping

Archaeology, a field deeply rooted in history and exploration, is undergoing a transformation with the integration of artificial intelligence (AI). AI, and specifically deep learning, has proven to be a powerful tool in automatically detecting archaeological sites and digitizing historic mapping. In this article, we will explore how AI is revolutionizing archaeology, its successful applications in site detection and mapping, as well as the challenges and future opportunities that this technology brings.

Introduction

Archaeology, the study of human history through the excavation and analysis of artifacts and structures, has traditionally relied on the expertise and intuition of archaeologists. However, with the advent of AI, the way we approach archaeological research is rapidly changing. By training AI models with known examples, such as Patterns, shapes, and features of archaeological sites, we can now teach computers to automatically detect and analyze these sites in large landscapes. This opens up exciting possibilities for the field, as it allows for scalable, efficient, and accurate site detection that was previously unimaginable.

Background and Training

To understand the power of AI in archaeology, it is essential to delve into the background and training methods used. Traditionally, archaeologists used pattern recognition techniques, manually instructing computers to look for specific shapes, edges, or lines that may represent archaeological sites. However, this approach had its limitations, as it required extensive manual input and was prone to errors and subjectivity. The introduction of deep learning revolutionized the field by allowing AI models to learn the features and patterns of archaeological sites on their own, without explicit programming.

My own background as an archaeologist and my fascination with the potential of automated approaches led me to pursue a Ph.D. focused on computer archaeology and deep learning. During my research, I discovered the effectiveness of training AI models with known examples of what we want to find. By providing the AI with examples of roundhouses, burial mounds, enclosures, and other archaeological features, it quickly learned to detect these sites with great accuracy.

The Power of AI in Archaeology

The power of AI in archaeology lies in its ability to Scale up site detection and provide valuable insights into previously undiscovered or overlooked archaeological sites. With the trained AI models, archaeologists can now extend their investigations to larger landscapes and cover more ground in a shorter amount of time. This scalability is particularly beneficial for organizations such as the Forestry Commission and the National Trust, as it allows them to detect region furrow, digitize historic maps, and preserve archaeological sites more effectively.

Detecting Archaeological Sites with AI

Pattern Recognition vs. Deep Learning

One of the key advancements in AI for archaeology is the shift from pattern recognition to deep learning. While pattern recognition involved manually instructing the AI on what to look for, deep learning enables the AI to learn the features and patterns internally. This means that instead of explicitly telling the AI to search for circles, lines, or edges, we provide it with known examples of the desired features, and it deduces these patterns on its own. The advantage of this approach is that the AI becomes more accurate and versatile in detecting archaeological sites, as it learns from a wider range of examples.

Examples of Successful Detections

The success of AI in archaeological site detection can be seen through various real-world examples. For instance, using deep learning, we were able to identify previously unknown roundhouses on the Isle of Aaron, which had been disregarded by archaeologists due to their overgrown state. By bounding these roundhouses and providing fresh eyes to examine them, we successfully brought them back into records and uncovered valuable historical information.

Furthermore, AI has proven effective in detecting deserted medieval villages, a unique pattern in the landscape that has no modern or ancient equivalents. By training the AI on a limited number of examples, we achieved high accuracy in detecting these villages. This breakthrough opens up new possibilities for understanding medieval settlements and exploring their historical significance.

Challenges and Limitations

While AI has shown great promise in archaeological site detection, there are challenges and limitations that need to be addressed. The availability of lidar data, which is crucial for accurate site detection, is still limited in some areas. However, the upcoming national coverage of lidar data holds great potential for scaling up AI applications in archaeology.

The AI's effectiveness in detecting certain features, such as cord rig, is also influenced by the availability of training examples. A wider range of examples and improved guidance from archaeological experts is needed to enhance the AI's accuracy in detecting these features. Additionally, the AI's reliance on patterns and the potential for confusion with modern equivalents pose further challenges that need to be overcome through research and refinement.

Scaling Up: From Local to National Coverage

The scalability of AI in archaeology is a significant advantage that allows for the extension of site detection from local to national coverage. By utilizing lidar data and national mapping programs, AI models can be trained and deployed across large areas, providing comprehensive site detection and preservation insights. The ability to cover vast landscapes in a shorter time frame enables archaeologists and organizations like the Forestry Commission and the National Trust to make informed decisions regarding land management, preservation, and future development.

Digitizing Historic Mapping with AI

In addition to site detection, AI has shown promising results in digitizing historic mapping, particularly in mapping orchards. By training AI models on 25-inch and 6-inch historic maps, we were able to automatically identify and map the location of orchards from the early 1900s to the 1950s. This automated approach offers a faster and more efficient alternative to manual mapping, providing valuable insights into the distribution and changes in orchards over time.

The digitization of historic mapping also offers opportunities for understanding biodiversity and land management practices. By analyzing the changes in Orchard distribution, we found a significant decline in small orchards since 1900, primarily replaced by farmland and impacted by urban expansion. These insights can inform land management strategies and conservation efforts, as well as contribute to our understanding of historical agricultural practices.

Next Steps and Research Opportunities

The successful applications of AI in archaeology and historic mapping open doors to further research and development. Future efforts will focus on expanding pilot studies, engaging in sector-wide discussions, and exploring the possibilities of integrating this technology into land development and construction practices. Moreover, there is immense research potential in refining AI models, enhancing accuracy, and discovering new insights through the analysis of slope, altitude, aspect, and other variables.

Collaboration among archaeologists, experts, and stakeholders is crucial in shaping the future of AI in archaeology and historic mapping. Roundtable Sessions and discussions will be organized to Gather insights, share best practices, and collectively explore the best ways to utilize this technology for the benefit of archaeological research, restoration, and land management.

Conclusion

The integration of AI in archaeology and historic mapping represents a paradigm shift in the field. By harnessing the power of deep learning, archaeologists can now detect and analyze archaeological sites more efficiently and accurately. The scalability of AI allows for comprehensive coverage, from local to national scales, enabling organizations and researchers to make informed decisions about land management, preservation, and development. Furthermore, the digitization of historic mapping provides valuable insights into changes in land use and biodiversity over time. As AI continues to evolve, the future of archaeology holds new possibilities and discoveries, transforming the way we understand and preserve our rich historical heritage.

Highlights

• AI is revolutionizing archaeology and historic mapping by enabling automated site detection and digitization.
• Deep learning has replaced traditional pattern recognition techniques, allowing AI models to learn features and patterns internally.
• AI has successfully detected unknown archaeological sites, such as roundhouses and deserted medieval villages.
• Lidar data availability and refined training examples are key factors in enhancing AI accuracy and scalability in site detection.
• AI-driven digitization of historic mapping offers insights into changes in land use, biodiversity, and orchard distribution over time.
• Collaborative efforts and sector-wide discussions are crucial in exploring the full potential of AI in archaeology and historic mapping.

FAQs

Q: What is the difference between pattern recognition and deep learning in AI for archaeology? A: Pattern recognition involves manually instructing the AI on specific shapes and features to look for, while deep learning allows the AI to learn these patterns internally without explicit programming.

Q: How accurate is AI in detecting archaeological sites? A: AI has shown high accuracy in detecting archaeological sites, especially when trained on a wide range of examples. However, challenges such as confusion with modern equivalents and limited lidar data availability can affect accuracy in certain cases.

Q: How scalable is AI in archaeology? A: AI enables scalable site detection by covering large landscapes and reducing the time required for manual investigations. As lidar data becomes more widely available, AI applications in archaeology can be scaled up from local to national coverage.

Q: Can AI help with digitizing historic mapping? A: Yes, AI has been successfully used to digitize historic maps, particularly in mapping features such as orchards. By training AI models on historic maps, the location and changes in orchards over time can be automatically identified and analyzed.

Q: What are the future opportunities and research potential in AI for archaeology and historic mapping? A: Future opportunities include refining AI models, enhancing accuracy, and exploring variables such as slope, altitude, and aspect for further insights. Sector-wide discussions and collaborations will be crucial in maximizing the potential of AI in these fields.

Q: How can AI contribute to land management and preservation? A: AI can provide valuable insights into archaeological site distribution, changes in land use, and biodiversity, allowing for informed decision-making in land management, preservation efforts, and future development plans.

Resources

• Geovation
• Royal Academy of Engineering
• Nesta
• UK Space Agency
• European Space Agency