Classification of soil types using convolution neural network
Hemasai Channareddy, Purvaja durga Barnala, Manjusha Kandukuri, Swarna Kuchibhotla · AIP conference proceedings · 2022
Soil classification is primarily concerned with systematic identification of soils dependent on differentiating properties and criteria that guide usage decisions. We can advise farmers on crop types, tillage methods, irrigation, and soil fertility by assessing soil characteristics, among many other factors that can affect crop production. One of many archaeological research topics is identifying components and nanostructures in bottom sediments of thin layers, as these attributes can disclose the details about its deposit from which they have been retrieved, along with its evolution and nature, and also hints about their relation to individual and social contexts or deformation processes over time. We designed a Deep Learning based system using Convolutional Neural Networks (CNN) for the classification of variety of old soil formations. The observed findings are positive, indicating that model offered may be applied to this classification algorithm successfully. The developed training model is used for calculation of number of distinct individual nanostructures in testing images, with an accuracy of 84 percent and a median error of 52 percent.