Identifying Clandestine Gravesites

For most scenarios, regardless of data quality, there is currently no push-button solution for automatically detecting burials with HSI and/or ALS data. The archaeology community has long used aerial remote for detecting and mapping subsurface archaeological. Recent publication points to success with new methods using HSI data, but these examples involve subsurface features that are large, rectilinear objects and present themselves as fairly obvious crop markings. Burials, on the other hand, tend to present themselves as a random blob of pixels, making it difficult for a photo interpreter to notice them, or to differentiate them from other truly random blobs throughout the data.

This research tested ground-, air- and satellite-based remote sensing technologies for identifying clandestine graves. Using known grave locations, including human graves interred specifically for this research project, CAST researchers joined the University of Tennessee-Knoxville in conducting field spectroscopy, aerial hyperspectral imaging (HSI), and terrestrial and aerial laser scanning (TLS and ALS) to model their surface changes over time.

Data was statistically compared within and between various field site data sets to determine the measure of reliable separability of individual disturbance targets (graves) in two different environments. Separate training and validation samples will determine whether new cases can be sorted correctly at rates better than random chance using the models developed from all the data collected. Existing models were developed using regression and discriminant function statistics. Improved models will be based on those approaches, but will also rely on additional statistical methods and neural networking.

Associated Grants and Awards

Increasing the Confidence and Reliability of Remote Sensing-based Predictive Models for Locating Gravesites, University of Tennessee (2016)