Comparison of two atmospheric correction methods for the classification of spaceborne urban hyperspectral data depending on the spatial resolution

Abstract

For remote-sensing applications such as spectra classification or identification, atmospheric correction constitutes a very important pre-processing step, especially in complex urban environments where a lot of phenomenons alter the shape of the signal. The objective of this article is to compare the efficiency of two atmo-spheric correction algorithms, COCHISE (atmospheric COrrection Code for Hyperspectral Images of remote-sensing SEnsors) and an empirical method, on hyperspectral data and for classification applications. Classification is carried out on several simulated spaceborne data sets with different spatial resolutions (from 1.6 to 9.6 m). Four classifiers are considered in the study: a k-means, a Support Vector Machine (SVM), and a sun/ shadow version of each of them, which processes sunlit and shadowed pixels separately. Results show that the most relevant atmospheric method for classification depends on the spatial resolution of the processed data set. Indeed, if the empirical method performs better on high-resolution data sets (up to 4%), its superiority fades out as the spatial resolution decreases, especially with the lower spatial reso-lution where COCHISE can be 10% more accurate than the empiri-cal method.

Hyperspectral Imagery for Environmental Urban Planning

“…-The comparison of classification performances from Pleiades, Sentinel 2 and 8m/30m GSD hyperspectral space sensor highlights the benefits of the SWIR spectral range but also the complementarity of the good classification between a high spatial and multispectral image and an 8m GSD hyperspectral acquisition [17]. In addition, it has been shown that using a subset of spectral bands well adapted to a classification problem makes it possible to obtain as good results as using the whole spectrum In return, experiments on the automatic determination of these well suited bands (position, width) for different case studies have been performed [18].…”

“…These experiments involve the ONERA Hyspex hyperspectral camera (GSD 1m [0.4, 1.0 µm] and 2 m GSD [1.0, 2.5 µm]), the IGN PAN CamV2 camera (20cm GSD).The Hyspex airborne acquisition is used to simulate at top of atmosphere the different space borne sensors with their related GSD in the different spectral bands. The images in radiance at sensor output are then corrected from the atmosphere using 2 different atmospheric correction methods as described in [17]. As we used the same airborne acquisitions to simulate the top of atmosphere images in the different spectral bands of each sensor, all the images will be correctly registered by construction.…”

Imagerie Hyperspectrale Et Zones Urbaines: Résultats Du Projet HYEP

“…On the other hand, Chen et al (2013) proposed an alternative by estimating the radiative components with the flat scene assumption and then by weighting them with empirical laws. (Roussel et al, 2018) conducted this comparison on the 2015 Toulouse dataset at different spatial resolutions by analyzing the classification performances using two classifiers: k-means (nonsupervised) and Support Vector Machine (SVM, supervised). The main results showed the best choice of the atmospheric corrections depends on the spatial resolution.…”

Comparison of Hyperspectral Techniques for Urban Tree Diversity Classification

“…This method assumes a flat ground hypothesis and estimates the water vapor content using a linear regression report (LIRR). It is particularly well suited for hyperspectral imagery [51].…”