Analysis Of The Parametrization Needs Of Different Land Cover Classifiers: The Case Study Of Granda Province (Spain)

Conference paper
Part of the Lecture Notes in Earth System Sciences book series (LNESS)

Abstract

Land cover monitoring and mapping is one of the main applications of the data provided by Earth Observing Satellites. The classification of extensive areas requires robust and operative methods. The most notable machine learning (ML) classification algorithms developed over the past years include: classification trees (CT), artificial neural networks (ANN), support vector machines (SVM) and Random Forest (RF). Although significant progress has been made in image classification based upon ML, a number of issues are yet to be resolved, particularly regarding parametrization. This paper discusses the limitations and crucial issues related to the application of different up-to-date ML classifiers. The results of four ML methods were quantitatively analyzed for the classification of land covers of a Mediterranean area, considering numerous parameter settings. Landsat-5 Thematic Mapper data captured in European spring and summer were used with auxiliary variables derived from a Digital Terrain Model to classify fourteen different land categories in the south of Spain. Overall, statistically similar accuracies of over 91 % were obtained for ANN, SVM and RF. The CT performed worse than the rest (overall accuracy of 86 %). However, the findings of this study show important differences in the efficiency of the classifiers, being RF the most accurate classifier with a very simple parametrization.

Keywords

Support Vector Machine Land Cover Machine Learning Artificial Neural Network Random Forest 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Notes

Acknowledgments

We are grateful for the financial support given by the Spanish MICINN (Project CGL2010-17629), Junta de Andalucía (Group RNM122) and the Portuguese Fundação para a Ciência e a Tecnologia (SFRH/BPD/89082/2012).

References

  1. 1.
    Congalton, R. G., & Green, K. (2009). Assessing the accuracy of remotely sensed data: Principles and practices (2nd ed.). Boca Raton, Florida: CRC Press.Google Scholar
  2. 2.
    Rodriguez-Galiano, V. F., & Chica-Rivas, M. (2012). Evaluation of different ML methods for land cover mapping of a Mediterranean area using multi-seasonal Landsat images and digital terrain models. International Journal of Digital Earth, doi: 10.1080/17538947.2012.748848.
  3. 3.
    Rodriguez-Galiano, V. F., Ghimire, B., Rogan, J., Chica-Olmo, M., & Rigol-Sánchez, J. P. (2012). An assessment of the effectiveness of a random forest classifier for land-cover classification. ISPRS-Journal of Photogramm Remote Sensing, 67, 93–104.CrossRefGoogle Scholar
  4. 4.
    Venables, W. N., & Ripley, B. D. (2002). Modern applied statistics with s. statistics and computing (4th ed.). New York, USA: Springer.Google Scholar
  5. 5.
    Yang, X. (2011). Parameterizing support vector machines for land cover classification. Photogrammetric Engineering and Remote Sensing, 77(1), 27–37.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2014

Authors and Affiliations

  • Víctor F. Rodriguez-Galiano
    • 1
  • Mario Chica-Olmo
    • 1
  1. 1.Dpto. de GeodinámicaUniversidad de GranadaGranadaSpain

Personalised recommendations