Mapping Urban Materials by Integrating Very High-Resolution Hyperspectral, Thermal, and LiDAR Data

Keywords: urban materials, very high resolution, hyperspectral imaging, thermal, LiDAR
Session Type: Virtual Paper Abstract
Day: Saturday
Session Start / End Time: 2/26/2022 11:20 AM (Eastern Time (US & Canada)) - 2/26/2022 12:40 PM (Eastern Time (US & Canada))
Room: Virtual 68

Abstract

Mapping detailed urban surface features has important implications on road and building maintenance, energy efficiency, city micro-climate, and urban sustainable design, etc. Yet collecting this information remains to be challenging due to the heterogeneity of the urban environment. This project intended to explore the capabilities of very high-resolution hyperspectral imaging (HI) by incorporating thermal and LiDAR data to characterize detailed urban landscape features based on the campus of the University of Northern Iowa in Cedar Falls, Iowa. A total of 231GB data were processed to derive different variables including 16 vegetation indexes, 5 minimum noise fraction bands, and 5 principal components bands from the hyperspectral image, while 6 topographic variables were extracted from the LiDAR data (e.g., surface roughness, hillshade, and intensity). Their relationships with thermal data were then examined using various approaches such as correlation analysis, fractal measure, as well as variation, skewness, and kurtosis statistics. Finally, around six different urban materials (metal, asphalt, concrete, etc.) were classified for the study area using the spectral angle mapper (SAM) algorithm. It was found that deriving precise and useful information from HI alone still remains to be a challenge. The incorporation of in situ and ancillary data, which may not always be available, is still required. The detailed interpretation products extracted from HI also call for new standards for result validation and accuracy assessment.