Soil Classification using Non-Imagery Spectroscopic Data

Plant development that is healthy relies on the minerals found in the soil. Conventional chemical analysis techniques can be used to retrieve information about soil nutrients. Non-imaging spectroscopy-based remote sensing methods offer a quick and non-destructive way to obtain data on soil nutrients. In this article, we attempt to classify soil samples based on pH, total nitrogen, and soil organic matter amounts. Spectral reflectance of the soil between 350 and 2500 nm is employed. To lessen the hyperspectral data's dimensionality, principal component analysis is applied. Subsequently, the primary components are used in decision tree and support vector machine classification of soil data. The findings demonstrate that soil organic matter can be classified with a 99.9% accuracy using a decision tree and an 81.45% accuracy using a support vector machine.