AI-Driven in Crop Disease Prediction and Management System

Abstract

Farming communities in developing countries face significant challenges including delayed soil testing, subjective crop selection, imbalanced fertilizer application, and late disease diagnosis. This paper presents a unified artificial intelligence system that integrates four essential agricultural modules into a single web platform. The system uses a deep learning optical character recognition pipeline to extract nitrogen, phosphorus, potassium, and pH values from uploaded soil report images. A Random Forest classifier then recommends the most suitable crop based on soil parameters and environmental data. Following Liebig\'s Law of the Minimum, a rule?based engine calculates nutrient deficits and solves a linear equation system to suggest precise quantities of DAP and urea fertilizers. Finally, a MobileNetV2 convolutional neural network classifies plant leaf images into disease categories and provides treatment advice. The frontend is built with React.js, the backend with Node.js and Express.js, and machine learning models are deployed as microservices using FastAPI. Experimental evaluation shows that the OCR module achieves 100% precision, the crop prediction model performs reliably across datasets, the fertilizer module yields balanced recommendations, and the disease detection model attains 88% validation accuracy. The system provides real?time, user?friendly assistance, reducing dependency on agricultural experts and enabling data?driven farming.

Introduction

The text describes the development of an AI-based smart agriculture system designed to help farmers make faster, more accurate decisions about soil health, crop selection, fertilizer use, and plant disease management. It highlights that traditional agricultural practices depend heavily on manual observation and laboratory testing, which are slow, costly, and often inaccurate, especially in rural areas.

Recent advances in machine learning and deep learning provide new solutions, but existing systems are usually limited to single tasks and are not integrated. To address this gap, the proposed system combines four key modules into one unified platform: OCR-based soil report digitization, crop prediction using Random Forest, rule-based fertilizer recommendation, and deep learning-based disease detection.

The system is built as a web application using a three-tier architecture with React.js for the frontend, Node.js/Express for the backend, MongoDB for data storage, and FastAPI microservices for ML models. Users can securely access the platform through OTP-based authentication.

Each module serves a specific function: OCR extracts soil nutrient values (N, P, K, pH) from uploaded reports using DBNet and CRNN models; the crop recommendation module uses a Random Forest classifier trained on environmental and soil features; fertilizer recommendations are generated using Liebig’s Law of the Minimum to compute nutrient deficiencies; and disease detection (based on CNN models like MobileNet/ResNet) identifies plant diseases from images.

Conclusion

This paper presented an AI?Based Smart Agriculture Assistant System that integrates soil analysis, crop prediction, fertilizer recommendation, and disease detection into a single web platform. The OCR module extracts soil nutrients from report images with perfect precision. The Random Forest model recommends suitable crops based on soil and environmental data. The rule?based fertilizer module calculates nutrient deficits and suggests optimal fertilizer amounts following Liebig\'s Law of the Minimum. The MobileNetV2 disease detection model identifies crop diseases from leaf images with 88% validation accuracy. The system is implemented using modern web technologies, provides real?time results, and offers a user?friendly interface. The system can be extended to support more crops and diseases by collecting additional training data. Real?time weather integration would improve crop prediction accuracy. A mobile application with offline capabilities would increase accessibility in remote rural areas. IoT sensors could provide live field data such as soil moisture and temperature. Multilingual support would help farmers from different linguistic regions. Advanced AI models like Vision Transformers could further boost disease detection accuracy.

References

[1] World Bank, \"What a Waste 2.0: A Global Snapshot of Solid Waste Management,\" World Bank Publications, 2018. [2] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, \"You Only Look Once: Unified, Real?Time Object Detection,\" in Proc. IEEE CVPR, 2016. [3] G. Jocher et al., \"Ultralytics YOLOv8,\" 2023. [Online]. Available: https://github.com/ultralytics/ultralytics [4] A. Krizhevsky, I. Sutskever, and G. Hinton, \"ImageNet Classification with Deep CNNs,\" in Proc. NIPS, 2012. [5] K. He et al., \"Deep Residual Learning for Image Recognition,\" in Proc. IEEE CVPR, 2016. [6] A. Howard et al., \"MobileNets: Efficient CNNs for Mobile Vision Applications,\" arXiv, 2017. [7] T. Lin et al., \"Microsoft COCO: Common Objects in Context,\" in Proc. ECCV, 2014. [8] Roboflow, \"Image Annotation Platform,\" [Online]. Available: https://roboflow.com [9] FastAPI, \"FastAPI Documentation,\" [Online]. Available: https://fastapi.tiangolo.com [10] MongoDB Inc., \"MongoDB Documentation,\" [Online]. Available: https://www.mongodb.com/docs [11] Facebook Inc., \"React.js Documentation,\" [Online]. Available: https://react.dev [12] Node.js Foundation, \"Node.js Documentation,\" [Online]. Available: https://nodejs.org [13] Express.js, \"Express Framework,\" [Online]. Available: https://expressjs.com [14] OpenCV, \"Open Source Computer Vision Library,\" [Online]. Available: https://opencv.org [15] S. Kaza et al., \"Global Waste Management Trends,\" World Bank, 2018.

Copyright

Copyright © 2026 Rutuja Margode, Khushi Kadu, Gokul Gole, Uday Takarkhede, Rahul Kanthale, Prof. Harish Gwalani, Prof. Samir Ubhale. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.