Data reduction of hyperspectral remote sensing data for crop stress detection using different band selection methods

The demand for sensor-based decision support in agriculture is rapidly growing which enhances precision of agricultural management. A fast and precise identification of fungal pathogen infections in crops is essential for the implementation of site-specific fungicide applications. Hyperspectral data collect spectral reflectance in contiguous bands over a broad range of the electromagnetic spectrum that allows examining stress-dependent shifting in certain spectral wavelengths caused by fungal infections. However, not all spectral information is needed for an accurate determination of pathogen infected areas. This study focuses on data reduction of hyperspectral image data for the identification of relevant and redundant information within the spectrum. By applying band selection techniques, different variants of fungal infected and vital wheat stands could be accurately differentiated and infected areas were localized.