Abstract for presentation at The 13th Australasian Remote Sensing and Photogrammetry Conference

Hyperspectral data can provide a number of quantitative information products for precision agriculture such as canopy cover, leaf area index, leaf chlorophyll and water content, soil organic matter, nitrogen, iron oxide and clay content. Although research over the last 15 years has produced many vegetation indices, leaf/canopy radiative transfer models, and hyperspectral image unmixing/classification techniques, operational use of such methods has largely been unsuccessful.
This paper demonstrates the use of spectroscopic analysis of hyperspectral data to derive robust indices for several plant and soil properties. The methods utilise the continuous nature of reflectance data with curve analysis techniques such as derivatives to measure changes in shape, position and depth of spectral features. These indices are sensitive to a wider range of variability in soil and plant variables. The effects of background soil and water on the crop reflectance are also removed by these indices. HyMap hyperspectral data over an irrigated agricultural area are used to demonstrate the crop/soil mapping with these new indices. The results show the detailed within field variability mapped by the indices. Another advantage of these new indices is that since they are based on unique physical/chemical properties of materials in an agricultural scene (crops, water, soil, stubble), they can also be used to classify the scene into various pure and mixed components. Examples are used to show a classification of the image based on these new indices alone, without any image unmixing/classification based on endmember spectra from the image or a spectral library. In summary, this paper demonstrates a complete end-to-end analysis of airborne hyperspectral data, where the entire processing sequence from atmospheric correction, image analysis and creation of indices, image georeferencing, mosaicking, to the final map products is based on the image and flight data, with no ancillary information required from the user.