Hyperspectral image feature extraction and classification for soil nutrient mapping

Aerial hyperspectral images were used for soil nutrient mapping and the image processing results were compared with the conventional field grid sampling and interpolation methods. A spatial low pass filter was applied to the hyperspectral imagery for enhancing soil nutrient property class separability. Image features were extracted from selective principal component transformed image space. Results showed that the supervised image classification could be implemented on a feature space with two features rather than on the original image space using all bands. It is concluded that using hyperspectral imagery for phosphorous and organic matter mapping could be a better approach than using the grid sampling and interpolation methods.