Paper Contents
Abstract
Soil structure affects agricultural productivity by affecting water retention, distribution of nutrients, and general crop health. Methods for traditional soil classification, which depend on laboratory analysis, are often slow and impractical for large-scale agriculture. To remove these boundaries, appoint our project Convolutional Neural Network (CNN) to classify the soil structure from the images taken in the uncontrolled Field Conditions (UFC), such as accounting for environmental variations such as light, background and moisture level. The approach involves shaping the 48 48 pixels and preparing ground images by training a CNN model to distinguish between different soil types, including black, red, clay, peat, yellow, and cinder. In addition, an integrated crop recommendation system detects the most appropriate crops based on the soil type. The model's performance is evaluated using accuracy, precision, recall and F1 score, which ensures high reliability in classification results. For ease of use, the system is distributed through a Django-based network interface so that users can upload soil images for real-time classification and crop recommendations. This solution provides a scalable and cost-effective alternative for traditional soil testing and reduces the difference between laboratory analysis and practical agricultural applications.
Copyright
Copyright © 2025 Ketha Rithika. This is an open access article distributed under the Creative Commons Attribution License.
