Improved classification of forest types using RapidEye satellite data with red edge band: A case study of Mitake, Gifu Prefecture
2013
Goto, S. (Gifu University (Japan). River Basin Research Center) | Awaya, Y.
Reflectance in the red edge band of the spectrum (680 to 730 nm) is strongly correlated with leaf chlorophyll concentration and is thus often used in vegetation indices. However, it has not previously been used to classify vegetation types. The RapidEye satellite, which was launched in August 2008, is equipped with multispectral imagery sensor and provides the red edge band. Here, we used two RapidEye images of Mitake, Gifu Prefecture, Japan taken on July 6, 2011 and November 30, 2011 to see if they could be used to classify vegetation types in a forest. The overall accuracy and Kappa coefficient of unsupervised and supervised classification of vegetation in the RapidEye images with red edge band were better than those obtained with RapidEye images with no red edge band. In addition, the accuracy of vegetation classified by the normalized difference red edge index (NDRE) was better than that classified by the normalized difference vegetation index (NDVI), which does not use the red edge band. We developed a new classification index, the DRe (difference of red edge), defined as Re_e-Rsub(710), where Rsub(710) is the reflectance at Band 4 (710 nm) and Re_e is the reflectance at 710 nm estimated from a linear equation expressing reflectance as a function of wavelength based on the reflectance at Band 3 (657.5 nm) and Band 5 (805 nm). The overall accuracy and Kappa coincident obtained with the DRe were 0.6602 and 0.5438, respectively, indicating that the DRe is sufficient for forest classification. Together, these results show that the red edge band has useful information for forest classification. We are presently testing the DRe's ability to classify vegetation in RapidEye images obtained in other seasons.
Show more [+] Less [-]AGROVOC Keywords
Bibliographic information
This bibliographic record has been provided by Agriculture, Forestry and Fisheries Research Information Technology Center