Abstract
This paper describes the variations of soil moisture and organic matter using underground image textural indices and hyperspectral signatures. The image textural indices were extracted from video image data by a real-time soil spectrophotometer (RTSS), and hyperspectral signatures were recorded on a portable field spectrophotometer and an AISA airborne sensor for the wheat-growing season at TUAT field. In explaining the variation of soil properties, the neural network (NN) showed consistently outperformed stepwise multiple liner regression (SMLR) and provided minimal prediction errors for spatial soil properties. The result showed that significant improvement had occurred and indicated integrated field spectral signatures with image texture would be an appropriate required in this type of analysis.
