Comparison of Hyperspectral Techniques for Urban Tree Diversity Classification

Abstract

This research aims to assess the capabilities of Very High Spatial Resolution (VHSR) hyperspectral satellite data in order to discriminate urban tree diversity. Four dimension reduction methods and two classifiers are tested, using two learning methods and applied with four in situ sample datasets. An airborne HySpex image (408 bands/2 m) was acquired in July 2015 from which prototypal spaceborne hyperspectral images (named HYPXIM) at 4 m and 8 m and a multispectral Sentinel2 image at 10 m have been simulated for the purpose of this study. A comparison is made using these methods and datasets. The influence of dimension reduction methods is assessed on hyperspectral (HySpex and HYPXIM) and Sentinel2 datasets. The influence of conventional classifiers (Support Vector Machine –SVM– and Random Forest –RF–) and learning methods is evaluated on all image datasets (reduced and non-reduced hyperspectral and Sentinel2 datasets). Results show that HYPXIM 4 m and HySpex 2 m reduced by Minimum Noise Fraction (MNF) provide the greatest classification of 14 species using the SVM with an overall accuracy of 78.4% (±1.5) and a kappa index of agreement of 0.7. More generally, the learning methods have a stronger influence than classifiers, or even than dimensional reduction methods, on urban tree diversity classification. Prototypal HYPXIM images appear to present a great compromise (192 spectral bands/4 m resolution) for urban vegetation applications compared to HySpex or Sentinel2 images.

Species-level classification of urban trees from worldview-2 imagery in Debrecen, Hungary: an effective tool for planning a comprehensive green network to reduce dust pollution

“…It is therefore reasonable to improve the methods that use cost-effective data sources; Persson et al (2018), for example, achieved good accuracy in tree species identification with freely available Sentinel-2 data. There is also a possibility that in unfortunate circumstances a large number of bands can provide more confusion than separability (Brabant et al, 2019).…”

“…As their role is irreplaceable, we need to pay particular attention to their assessment. Moreover, the mapping of tree species that produce potentially allergenic pollen may prove beneficial as well (Brabant et al, 2019). Apart from biodiversity and tree health assessment, hazard and stress management, and monitoring of invasive species are also common research objectives (Fassnacht et al, 2016;Hüse et al, 2016).…”

Partial Linear NMF-Based Unmixing Methods for Detection and Area Estimation of Photovoltaic Panels in Urban Hyperspectral Remote Sensing Data

“…The full used hyperspectral image covers a part of Toulouse (France) urban area (Figure 3). Toulouse downtown is typical from old cities situated in the South of France with a large number of components/characteristics [32,33]: tile roofs, low rise homes, big architectural monuments (cathedral, town hall, etc. ), dense urbanization, small streets, large avenues and green spaces (public gardens, squares, etc.).…”

Mapping Succession in Non-Forest Habitats by Means of Remote Sensing: Is the Data Acquisition Time Critical for Species Discrimination?

“…For years, many scientists throughout the world, and especially those dealing with forestry, have undertaken research work on delineating various forest types or stands e.g., [31][32][33] and tree species e.g., [34][35][36][37][38][39][40][41][42][43][44][45][46][47]. The mapping of urban tree species has also been a widely studied research topic in the last decade e.g., [48][49][50][51][52][53][54]. A wide review of studies on the classification of tree species from remotely sensed data was presented by Fassnacht et al [55] and demonstrated that the number of studies focusing on the classification of tree species has been constantly increasing over the last four decades.…”

“…In a summary of their review, Fassnacht et al [55] found that passive optical systems, and especially hyperspectral systems, generally showed higher potential for the classification of tree species than active SAR or LiDAR (Light Detection and Ranging) sensor systems, as they enabled to obtain similar accuracies but for a higher number of tree species. In recent years, many studies have also examined the use of combination of LiDAR and hyperspectral data for identification of tree species [36,45,46,48,51,54,56]. The accuracy of tree species classification reported by different researchers varies.…”