Hyperspectral clustering and unmixing of satellite imagery for the study of complex society state formation.

详细信息   

• 作者：Kwong ; Justin.
• 学历：Master
• 年：2009
• 导师：Messinger, David W.,eadvisorKerekes, John P.ecommittee memberMiddleton, William D.ecommittee memberVodacek, Anthonyecommittee member
• 毕业院校：Rochester Institute of Technology
• Department：Imaging Science
• ISBN：9781109575231
• CBH：1473597
• Country：USA
• 语种：English
• FileSize：25544683
• Pages：80

文摘

This project is an application of remote sensing techniques to the field of archaeology. Clustering and unmixing algorithms are applied to hyperspectral Hyperion imagery over Oaxaca, Mexico. Oaxaca is the birthplace of the Zapotec civilization, the earliest state-level society in Mesoamerica. A passionate debate is ongoing over whether the Zapotecs' evolution was environmentally deterministic or socioeconomic. Previous archaeological remote sensing has focused on the difficult tasks of feature detection using low spatial resolution imagery or visual inspection of spectral data. This project attempts to learn about a civilization on the macro level, using unsupervised land classification techniques. Overlapping 158 band Hyperion data are tasked for approximately 30,000 km2, to be taken over several years. K-means and ISODATA are implemented for clustering. MaxD is used to find endmembers for stepwise spectral unmixing. Case studies are performed that provide insights into the best use of various algorithms. To produce results with spatial context, a method is devised to tile long hyperspectral flight lines, process them, then merge the tiles back into a single coherent image. Google Earth is utilized to effectively share the produced classification and abundance maps. All the processes are automated to efficiently handle the large amount of data. In summary, this project focuses on spectral over spatial exploitation for a land survey study, using open source tools to facilitate results. Classification and abundance maps are generated highlighting basic material spatial patterns (e.g., soil, vegetation and water). Additional remote sensing techniques that are potentially useful to archaeologists are briefly described for use in future work.