AI-IoT-Enabled Crop Monitoring Through Crop Stage and Leaf Disease Identification Using PECFIS and DGBESCNN

Abstract

The aspect of crop monitoring takes into consideration the timely detection of crop stages, leaf disorders, and deficiencies to enhance crop yield and decrease losses in agriculture. However, most of the current methods are limited to either disease detection or nutrient evaluation and do not examine the conditions of crops at various stages of growth, even though several AI -IoT-based solutions have been suggested to be applied to crop health monitoring. In addition, the estimation of the severity of the diseases is neglected, and this restricts decision-making in favor of the farmers. To address these constraints, the paper presents a Parametrized Elliptical Cauchy Fuzzy Inference System (PECFIS) combined with a Deep Glorot Bessel Elliott Softplus Convolutional Neural Network (DGBESCNN), proposed as an AI-based solution for crop monitoring and IoT support. The IoT devices in the form of drones are used to get real-time field images, and they are preprocessed in terms of noise reduction, contrast enhancement by LHM-CLAHE, conversion to HSV color space, and feature discrimination by vegetation indexing, as well as C3MEK-Means. PECFIS is used to determine eight key stages of rice growth and the severity of leaf diseases, whereas DGBESCNN provides proper classification of leaf diseases and nutrient deficiencies at each growth stage. The evaluation of the proposed framework was conducted using publicly available datasets on rice leaf disease and nutrient deficiency. The results of the experiments show that the system achieves high classification performance, with an accuracy of 98.82, a precision of 98.65, a recall of 98.73, an F1-score of 98.59, and low error rates (MSE = 0.0135, RMSE = 0.116). The findings show that the developed AI-IoT system is superior to available approaches and can serve as a dependable, real-time, and scalable solution in precision agriculture and intelligent crop monitoring.

Plant phenotyping relevance

ドローン画像からイネの生育段階と葉病害の重症度を推定・分類するAI-IoT手法が研究の中心であり、植物状態の取得・抽出方法を技術的に評価している。

Code and data availability

The paper evaluates its PECFIS-DGBESCNN crop monitoring framework on two publicly available Kaggle datasets (Nutrient Deficiency Symptoms in Rice, 1,156 images; Rice Leaf Diseases, 120 images), with explicit dataset links provided by the authors. No author code, models, or other paper-specific assets are shared.