Impervious surfaces, including rooftops, pathways and parking lots, are surfaces through which water cannot permeate and are recognized as significant features of urban environments. This project aims to classify impervious urban materials at the University of Northern Iowa (UNI), Iowa, USA, by using very high resolution Airborne Imaging Spectrometer for Applications (AISA) hyperspectral imagery and Forward-Looking Infra Red (FLIR) Phoenix imagery. The classification divided impervious surface materials into six main classes: roof concrete, road concrete, shingle, metal, EPDM (which is a rubber material widely used in low-slope roofing throughout the USA and worldwide; EPDM Roofing Association, 2017) and clay. In order to analyze AISA data, the Quick Atmospheric Correction (QUAC) was applied using Environment for Visualizing Images (ENVI) software to eliminate atmospheric attenuation. Image-to-image registration was used to match the hyperspectral image with a reference image. ENVI’s Maximum Likelihood was used to classify the imagery. The overall accuracy using only hyperspectral imagery was 59%. The addition of a thermal band eliminated most inaccuracy caused by shadows and raised the overall classification accuracy to 73%. Some complex real life variables, such as 1) cars, 2) isolated misclassified pixels (noise), as well as 3) discrepancies between classification based on UNI building survey and materials exposed to air in real life caused by lack of site visit prevented the classification algorithm from obtaining a higher accuracy. In addition, radiance values for classes were calculated in order to relate to the Urban Heat Island (UHI) effect. Unlike expectation, that EPDM should have the highest radiance due to its dark color, results showed that road concrete and vegetation have the highest radiances, meaning they have highest temperatures and contributed to the UHI effect most on the UNI campus. The phenomenon may be explained by air conditioning inside the buildings that cools building rooftops down. Air conditioners need be turned off when taking the thermal image to obtain more reliable radiance values to relate to UHI effect.
Wu, Feiyi, "Urban Surface Material Mapping Using Very High Resolution Hyperspectral Imagery and Thermal Imagery" (2018). Senior Independent Study Theses. Paper 8018.
remote sensing, hyperspectral imagery, thermal imagery, material classification, impervious surfaces
Bachelor of Arts
Senior Independent Study Thesis
© Copyright 2018 Feiyi Wu