Detecting and characterizing nighttime lighting using multispectral and hyperspectral imaging

Abstract

Multispectral imagery (MSI) of Las Vegas, Nevada, were investigated to determine their potential for accurately mapping nocturnal lighting using a reference spectral library of lighting types. Nocturnal lighting classifications of International Space Station (ISS) astronaut color photography and 8-, 6-, and 4-band MSI generated by modeling high spectral resolution hyperspectral imagery (HSI) data to lower spectral resolution were compared to lighting identification accomplished using the full resolution HSI spectral signatures. The results indicate that ISS imagery does not have the spectral resolution necessary to accurately distinguish between the emission features of outdoor lighting. The modeled multispectral band configurations demonstrated somewhat improved separation of certain lighting types by their spectral signatures, however, with only 8, 6, or 4 spectral bands, accurate discrimination of lighting types still remains a daunting task. While the different colors associated with outdoor lighting can be visually delineated in MSI data, the limited spectral information does not allow for accurate lighting type classification because of the inability to identify specific sharp emission features. Mapping nocturnal lighting using MSI data does have some utility, and certainly would provide better spatial coverage, however, HSI remains the most accurate method to differentiate the emission features associated with urban outdoor lighting.