MULTI-TEMPORAL SPECTRAL UNMIXING FOR SCALE-AWARE LAND USE/LAND COVER CLASSIFICATION USING SENTINEL-2

Download

LeiliNABDEL_PhDthesis.pdf

Date

2022-9-02

Author

NABDEL, Leili

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

127
views

87
downloads

CORINE Land Cover mapping scheme provides an ontology to create consistent classification maps that are regularly updated. Since CORINE is a mixed Land Use / Land Cover (LULC) standard, classification based solely on spectral and temporal properties of the pixel in question is not adequate and contextual information should also be considered. An efficient classification method, which uses spectral unmixing of multi-spectral multi-temporal Sentinel-2 imagery, has been proposed in this work. Rather than the classical pixel-based classification approach, the task is considered as a scene classification problem, where the scene is defined as the neighborhood of the pixel to be classified. Since the pixels may contain multiple materials, it is difficult to interpret the measured spectra. Hence, the scene is analyzed by first estimating the abundances of endmembers of the basic materials generated via supervised spectral unmixing. The abundances of basic materials, their distribution in the neighborhood, and their temporal properties are characteristic to classes; hence the chosen features reflect these characteristics. The random forest algorithm is then used for classification. Two different ground truth sets are created for each of the two study areas to see the effect of different sampling strategies. The overall accuracy for the first study area in Seydişehir, Turkey is 86.3%, 97.4%, or 72.6%, depending on the sampling strategy for the fourth CORINE hierarchical level. Meanwhile, the overall accuracy for the second area in Rüdersdorf, Germany is 75.0%, 99.4%, or 85.9% for the third CORINE hierarchical level. For better evaluation of the capabilities of the proposed method for larger areas, the performance of generalized area is also tested. Obtained results, specified that the proposed approach outperforms the typically classification methods specially in deeper hierarchy levels.

Subject Keywords

Spectral Unmixing, Land Cover Classification, Random Forest, Multi-Temporal Endmembers, CORINE

URI

https://hdl.handle.net/11511/99485

Collections

Graduate School of Natural and Applied Sciences, Thesis

Suggestions

OpenMETU
Core

Evaluation of the NDVI in plant community composition mapping: a case study of Tersakan Valley, Amasya County, Turkey DOĞAN, Hakan Mete; Celep, Ferhat; Karaer, Fergan (Informa UK Limited, 2009-01-01) Mapping the composition of plant community types requires reliable spatial data obtained from field surveys and satellite-derived indices. The normalized difference vegetation index (NDVI) is the simplest and most frequently used index in plant applications. If relationships between the NDVI and plant cover abundance are determined, this information can be used in the mapping process. In this study, we investigated these possible connections for mapping the plant community composition of Tersakan Valley in ...
A Theoretical Analysis of Multi-Modal Representation Learning with Regular Functions Vural, Elif (2021-01-07) Multi-modal data analysis methods often learn representations that align different modalities in a new common domain, while preserving the within-class compactness and within-modality geometry and enhancing the between-class separation. In this study, we present a theoretical performance analysis for multi-modal representation learning methods. We consider a quite general family of algorithms learning a nonlinear embedding of the data space into a new space via regular functions. We derive sufficient condit...
Harmonicity of maps between (indefinite) metric-f-manifolds and rho-pseudo harmonic morphisms Erdem, S (2003-10-01) Conditions are investigated for maps to be harmonic between M(P)-f-manifolds with (semi-)Riemannian metrics. Also a geometrical condition, phi(r)-pseudo horizontally weak conformality [which is weaker than horizontally weak conformality when they are comparable], is imposed on maps of a (semi-) Riemannian manifold into a metric M(P)-f-manifold with (semi-) Riemannian metric and harmonicity of such maps are discussed. Some results obtained by Loubeau and Lichnerowicz in the almost Hermitian case are given he...
Deep Spectral Convolution Network for Hyperspectral Unmixing Akar, Gözde (2018-10-10) In this paper, we propose a novel hyperspectral unmixing technique based on deep spectral convolution networks (DSCN). Particularly, three important contributions are presented throughout this paper. First, fully-connected linear operation is replaced with spectral convolutions to extract local spectral characteristics from hyperspectral signatures with a deeper network architecture. Second, instead of batch normalization, we propose a spectral normalization layer which improves the selectivity of filters b...
Extended self organizing feature map: A tagged potential field approach Baykal, Nazife (KLUWER ACADEMIC PUBL, SPUIBOULEVARD 50, PO BOX 17, 3300 AA DORDRECHT, NETHERLANDS, 1999-8) This paper proposes an escape methodology to the local minima problem of self organizing feature maps generated in the overlapping regions which are equidistant to the corresponding winners. Two new versions of the Self Organizing Feature Map are derived equipped with such a methodology. The first approach introduces an excitation term, which increases the convergence speed and efficiency of the algorithm, while increasing the probability of escaping from local minima. In the second approach, we associate a...

Citation Formats

L. NABDEL, “MULTI-TEMPORAL SPECTRAL UNMIXING FOR SCALE-AWARE LAND USE/LAND COVER CLASSIFICATION USING SENTINEL-2,” Ph.D. - Doctoral Program, Middle East Technical University, 2022.