Scholarlink

<p>This work presents the development and deployment of a lightweight, embedded machine learning-based microsystem for real-time prediction of primary soil macronutrients-Nitrogen (N), Phosphorus (P), and Potassium (K). The system integrates a multi-parameter soil sensor with a Raspberry Pi 5 platform, leveraging inputs such as Electrical Conductivity (EC), temperature, pH, and humidity to estimate nutrient levels accurately. Proximate analysis of multiple machine learning algorithms-including Random Forest Regressor (RFR), Gradient Boosting Regressor (GBR), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Linear Regression (LR)-was performed, achieving an <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(R^2\)</EquationSource> </InlineEquation> &gt;0.995 for nutrient prediction. The trained model is deployed on-device using a FastAPI-based inference engine, enabling edge-level predictions without reliance on cloud infrastructure. A Node-RED dashboard is also integrated for real-time visualization of EC values and predicted NPK levels. The proposed solution offers a cost-effective, scalable, and field-deployable tool for precision agriculture, empowering farmers with actionable soil health insights in resource-constrained environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Embedded microsystem for soil nutrient estimation using machine learning models

  • 期刊论文

  • 发表日期：

    2025年9月1日

查看全文
  • Sneha Dattatreya,
  • Archit Khurana,
  • Kanjalochan Jena,
  • Gaurav Chatterjee

摘要

This work presents the development and deployment of a lightweight, embedded machine learning-based microsystem for real-time prediction of primary soil macronutrients-Nitrogen (N), Phosphorus (P), and Potassium (K). The system integrates a multi-parameter soil sensor with a Raspberry Pi 5 platform, leveraging inputs such as Electrical Conductivity (EC), temperature, pH, and humidity to estimate nutrient levels accurately. Proximate analysis of multiple machine learning algorithms-including Random Forest Regressor (RFR), Gradient Boosting Regressor (GBR), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Linear Regression (LR)-was performed, achieving an \(R^2\) >0.995 for nutrient prediction. The trained model is deployed on-device using a FastAPI-based inference engine, enabling edge-level predictions without reliance on cloud infrastructure. A Node-RED dashboard is also integrated for real-time visualization of EC values and predicted NPK levels. The proposed solution offers a cost-effective, scalable, and field-deployable tool for precision agriculture, empowering farmers with actionable soil health insights in resource-constrained environments.