Most available hyperspectral image processing algorithms analyse the data based on the spectral information exclusively and do not treat the data as an image. Methods that use spatial and spectral information mostly perform a two-step processing technique, i.e. the spatial analysis is used as a retrospective smoothing. In this paper a recently developed method, which is based on the beamlet analysis, is presented. This method, which uses spectral and spatial information simultaneously, was adjusted in this study to multi- and hyperspectral images. The results of the new methodology, called the recursive dyadic partitioning =beamlet decorated (RDP-BD), were compared to fuzzy C-means spectral classification and to spatially coherent regions classification. These methods were used to classify nitrogen levels in an experimental potato plot, which was grown under different nitrogen treatments. The image was classified by these three methods based on the combination of two spectral indexes: transformed chlorophyll absorption reflectance index (TCARI) and optimized soil-adjust vegetation index (OSAVI), the first three principal components of the image and the complete 210 bands of the hyperspectral image. Although the spatial analysis slightly improved the classification (from71% to 78%), the most pronounced contribution was its ability to divide the experimental plot into zones that fit the treatments borders. In addition, it was shown that the principle components and the whole spectra out-performed the spectral index. These methods should assist with testing the efficiency of variable rate nitrogen fertilisation as they utilise the informative spectral data to properly classify nitrogen level in the pixel scale and delineate management zones at the field scale.
|Number of pages||9|
|State||Published - 2013|
Bibliographical noteFunding Information:
This project was supported by the Israeli Space Agency , Israeli Ministry of Science and also by Research Grant Award No. CA-9102-06 from BARD-AAFC – The United States–Israel Binational Agricultural Research and Development Fund and Agriculture and Agri-Food , Canada.