An effective model for predicting agricultural crop yield on remote sensing hyper-spectral images using adaptive logistic regression classifier

Abstract

In today's world, agriculture can be a major significant source of endurance as well as an essential aspect in the development of the world economy. Hyper-spectral (HS) imaging plays a vital role in remote sensing to support the agriculture environment where the satellite communication technology is utilized to provide improved accessibility of HS imagery demands the assessment of precision agriculture application. The historical statistical information may not be helpful in the prediction of agriculture yield. Therefore, in this research, a novel HS image processing technique has been implemented using remote sensing technology to achieve an efficient and accurate prediction of crop yield. Primarily, the preprocessing technique called 2D-adaptive anisotropic diffusion filter is implemented to eradicate the speckle noise from HS images and then enhanced by applying a new edge preservation-contrast limited adaptive histogram equalization technique. Afterward, the fast fuzzy C means algorithm has been incorporated in the proposed method to cluster the agriculture crop's gradient intensity pixels from HS images. Finally, the optimum features are learned and then fed into the adaptive logistic regression classifier to classify the different agricultural crops. The simulation results manifest that the proposed methodology efficiently predicts the agricultural crops with superior accuracy of 99.27% and specificity of 99% as compared with existing state-of-the-art techniques.