Agriculture is considered as one of the indirect sustainable development goals leading towards life on land and zero hunger. To obtain better yields from agriculture, it is important to maintain soil conditioning. The most important soil conditioning is the manuring of the field well before cultivation. Hyperspectral remote sensing images, which has several applications in agriculture domain, can be utilized for the classification of manure usage. A set of features must be extracted from hyperspectral images including derived spectral indices (such as vegetation indices) and then to be passed through machine learning algorithms for classification. Although several hyperspectral indices are available, they are not well suited for all types of classification, including manure usage. Hence, in this paper a hyperspectral index is evolved using a ternary particle swarm optimization for the manure usage classification. The results of manure usage classification in pasture fields with the evolved hyperspectral index using convolutional neural network are promising when compared to those of other hyperspectral indices.
Reghunadhan et al. (Mon,) studied this question.