Page 682 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

(continued)

Item Details
The approach tackles these challenges by integrating advanced remote sens-
ing data (from Synthetic Aperture Radar (SAR) and high-resolution multispectral
imagery) with machine learning and geospatial analysis techniques. The goal is to
automate the detection process and improve guidance for future archaeological
investigations at sites like Saruq Al-Hadid in Dubai, UAE.
Key elements of the proposed solution include:
Multi-Sensor Remote Sensing Integration: Utilization of high-resolution multispec-
tral imagery (Worldview-3) and SAR data (Advanced Land Observing Satellite
(ALOS)-2/Phased Array type L-band Synthetic Aperture Radar (PALSAR)-2) to
enhance surface and subsurface feature detection.
Advanced Image Processing and Feature Enhancement: Application of Principal
Component Analysis (PCA), spectral band transformations, and band ratio indices
(e.g., Normalized Difference Vegetation Index (NDVI), Normalized high-frequency
difference (NHFD)index to enhance and extract meaningful features from the
remote sensing data.
Application of artificial intelligence: Implementation of machine and deep
learning algorithms—including unsupervised classification (e.g., Iterative Self-Or-
ganizing Data Analysis Technique Algorithm (ISODATA)), K-means++ clustering)
Key Aspects and convolutional neural networks (CNNs) for semantic segmentation to auto-
of Solution matically detect features like dune patterns and potential archaeological features.
Geospatial Analysis and Data Integration: Extraction and processing of geospatial
data (e.g., digital elevation models, slope maps, hydrographic networks and dune
morphology) combined with geostatistical methods to analyze spatial patterns
and relationships among various landscape elements.
Multimodal Reclassification and Expert Input: Fusion of remote sensing data,
geospatial analysis, and field expertise to reclassify the landscape into distinct
geomorphological assemblages that distinguish between natural formations and
potential archaeological sites.
Predictive Modeling and Pattern Recognition: Development of predictive models
using techniques such as Kernel PCA and Gaussian radial basis function (RBF)
kernels to identify and prioritize areas of potential archaeological significance,
guiding future on-site investigations.
Validation through field data, excavation reports, and supporting geophysical
surveys—including Ground Penetrating Radar (GPR) and magnetic survey—helped
confirm the accuracy of the predictions.
These components collectively facilitate more efficient and automated archae-
ological prospection.

Remote Sensing, Synthetic Aperture Radar (SAR), Multispectral Imagery (Worl-
dview-3, ALOS-2/PALSAR-2), Spectral Band Transformation, Band Ratio Indices
(NDVI, NHFD), Principal Component Analysis (PCA), Advanced Image Processing,
Unsupervised Classification (ISODATA, K-means++), Machine Learning, Deep
Technology Learning, Convolutional Neural Networks (CNNs), Gaussian RBF Kernels, Geospa-
Keywords
tial Analysis, Digital Elevation Models (DEMs), Geographic Information Systems
(GIS), Archaeological Site Detection, Pattern Recognition, Data Integration, Data
Fusion, Ground Penetrating Radar (GPR), Magnetic Survey, Geophysical Valida-
tion.

Data Avail-
ability Public

646