Submitted to: Proceedings of the American Society of Agricultural and Biological Engineers International (ASABE)
Publication Type: Proceedings
Publication Acceptance Date: July 30, 2006
Publication Date: August 30, 2006
Citation: Yang, C., Everitt, J.H., Bradford, J.M. 2006. Use of spectral angle mapper (SAM) and hyperspectral imagery for yield estimation. Proceedings of the American Society of Agricultural and Biological Engineers International (ASABE). CDROM.
Interpretive Summary: Hyperspectral imagery contains nearly continuous spectral data and has the potential for better differentiation and estimation of biophysical attributes of interest. This study applied the spectral angle mapper technique to airborne hyperspectral imagery for estimating grain sorghum yield variability. Results show that grain yield was significantly related to spectral angle values and that spectral angle images derived from the hyperspectral imagery had better correlations with yield than majority of the normalized difference vegetation indices derived from the imagery. This technique provides a useful tool to convert a hyperspectral image to a single layer image to characterize relative yield variability without using actual yield data.
Vegetation indices (VIs) derived from remotely sensed imagery are commonly used to estimate crop yields. Spectral angle mapper (SAM) provides an alternative approach to quantifying the spectral differences among all pixels in imagery and therefore has the potential for mapping yield variability. The objective of this study was to apply the SAM technique to airborne hyperspectral imagery for mapping yield variability. Airborne hyperspectral imagery was acquired from two grain sorghum fields in south Texas and yield data were collected using a grain yield monitor. SAM was used to generate spectral angle images from the hyperspectral imagery. Statistical analysis showed that yield was significantly related to the SAM images. For comparison, all 5,151 possible normalized difference vegetation indices (NDVIs) were derived from the 102-band images and related to yield. Results showed that SAM provided higher r-values than 80% and 95% of the 5,151 NDVIs for fields 1 and 2, respectively, but the best NDVIs had better correlations with yield. Nevertheless, SAM provides a useful tool to convert a hyperspectral image to a single layer image to characterize yield variability without using actual yield data, while the best NDVIs can only be identified based on actual yield data. These results indicate that the SAM technique can be used alone or in conjunction with other VIs for yield estimation from hyperspectral imagery.