Soil health analysis and fertilizer prediction for crop image identification by Inception-V3 and random forest

Abstract

Pests and diseases are some of the most severe hazards to crops across the globe. Their presence may have negative implications for health, the economy, the ecosystem, and humanity. Many challenges relating to crop yield enhancement develop due to a lack of or lack of information, common in the agricultural industry. For this reason, it is necessary to create feasible models that take into account a variety of climate, soil, and plant characteristics to study the crops and make effective responses to increase production possible. In this work, classification of leaf disease was performed by utilizing Inception v3 methodologies; in addition, detection of soil fertility was done by employing Random Forest (RF).Leaf smut, Brown Spot, and Bacterial leaf blight are rice crop diseases that we'll discuss. We use an image sensor to snap diseased rice varieties in a rice field. The proposed work tests the efficacy of four different methods for removing the backgrounds and three other ways for segmenting data. The segmentation of the disease region of a leaf picture using InceptionV3 is what we recommend for getting the most accurate feature extraction possible. We improve the Random Forest (RF) output by predicting disease-specific fertilizer. Various aspects are categorized into colour, form, and texture by humans. In terms of accuracy, the model outperformed different other approaches, according to the results of the experiments. More than 98 percent of diseases predict accurately. This research has also shown an economical method of detecting plant diseases.In comparison with the prevailing benchmark algorithm, the proposed model achievesan accuracy and precision of 97% and 96%. So, it is evident that better performance is attained by the proposed methodology than the prevailing models.