Using GeoAI and Mixed-Data to Classify Built Heritage

Abstract

Built heritage, which comprises structures that are valued for their historical or architectural characteristics, is important to community’s identity, sense of place, and, in many cases, economy. Many communities in Ontario have used Heritage Act provisions to protect locally significant buildings through local heritage registers or local bylaws that officially designate them as heritage structures. Recent changes introduced by Ontario’s Bill 23, the 2022 More Homes Built Faster Act, have significantly altered the heritage designation process by limiting how long a property can remain on a local heritage register. This change highlights the need for a more efficient method for municipalities to identify and classify potential heritage properties. Geospatial Artificial Intelligence (GeoAI), an innovative approach integrating AI with Geographic Information Science (GIS), has potential to automate heritage identification and classification tasks, and assist heritage planners. This thesis explores the potential of GeoAI to streamline heritage property identification and designation through three main models. These models leverage Convolutional Neural Networks (CNN) and Multi-Layer Perceptrons (MLP) to classify architectural styles, identify potential heritage properties, and predict heritage designations. The models were trained on non-spatial and geospatial datasets, including archival photographs and region-specific data from Ontario, enhancing their ability to detect architectural details and heritage features unique to the area. The results demonstrate the effectiveness of these models, with the Architectural Style Classification Model achieving a 89% accuracy, despite challenges with similar styles. The Heritage Identification Model significantly improved efficiency with a 96.62% accuracy rate, while the Heritage Property Designation Prediction Model, combining CNN and MLP approaches, achieved 96% accuracy. The findings highlight the potential of AI and GeoAI to aid heritage practices with new technological methods. This research contributes to the broader knowledge base by providing refined tools for decision-making in heritage conservation and also suggests directions for future research to further optimize the integration of GeoAI in heritage tasks.