Object- and pixel-based analysis for mapping crops and their agro-environmental associated measures using QuickBird imagery

  • Authors:

  • Isabel Luisa Castillejo-González;Francisca López-Granados;Alfonso García-Ferrer;José Manuel Peña-Barragán;Montserrat Jurado-Expósito;Manuel Sánchez de la Orden;María González-Audicana

  • Affiliations:

  • Department of Cartography Engineering, Geodesy and Photogrammetry, University of Córdoba, Campus de Rabanales, 14071 Córdoba, Spain;Institute for Sustainable Agriculture, CSIC, Apdo. 4084, 14080 Córdoba, Spain;Department of Cartography Engineering, Geodesy and Photogrammetry, University of Córdoba, Campus de Rabanales, 14071 Córdoba, Spain;Institute for Sustainable Agriculture, CSIC, Apdo. 4084, 14080 Córdoba, Spain;Institute for Sustainable Agriculture, CSIC, Apdo. 4084, 14080 Córdoba, Spain;Department of Cartography Engineering, Geodesy and Photogrammetry, University of Córdoba, Campus de Rabanales, 14071 Córdoba, Spain;Department of Projects and Rural Engineering, Public University of Navarre, Campus de Arrosadía, 31006 Pamplona, Spain

  • Venue:
  • Computers and Electronics in Agriculture
  • Year:
  • 2009

Quantified Score

Hi-index 0.01

Visualization

Abstract

Currently, monitoring of agrarian policy actions usually requires ground visits to sample targeted farms, a time-consuming and very expensive procedure. To improve this, we have undertaken a study of the accuracy of five supervised classification methods (Parallelepiped, Minimum Distance, Mahalanobis Classifier Distance, Spectral Angle Mapper and Maximum Likelihood) using multispectral and pan-sharpened QuickBird imagery. We sought to verify whether remote sensing offers the ability to efficiently identify crops and agro-environmental measures in a typical agricultural Mediterranean area characterized by dry conditions. A segmentation of the satellite data was also used to evaluate pixel, object and pixel+object as minimum information units for classification. The results indicated that object- and pixel+object-based analyses clearly outperformed pixel-based analyses, yielding overall accuracies higher than 85% in most of the classifications and exhibiting the Maximum Likelihood of being the most accurate classifier. The accuracy for pan-sharpened image and object-based analysis indicated a 4% improvement in performance relative to multispectral data.