Abstract :

Soil Type Identification Is A Fundamental Requirement In Agriculture, Land-use Planning, Irrigation Design, And Environmental Management, Because The Physical And Chemical Nature Of Soil Determines Which Crops Can Be Grown, How Water Must Be Managed, And How Land Should Be Conserved. Conventional Identification Depends On Laboratory Assay And Expert Field Interpretation, Which Are Slow, Expensive, And Difficult To Scale Across Large Or Remote Regions. This Project Proposes An Image-based Soil-classification System Built On A Deep Convolutional Neural Network (CNN) That Predicts The Soil Type Directly From An Ordinary Camera Photograph, Removing The Need For Chemical Testing. The System Is Implemented In Python Using TensorFlow 2.10 And Keras For Model Development, NumPy And Pandas For Data Handling, Matplotlib And Seaborn For Visualisation, And A Streamlit Web Interface For Image Upload, Prediction, And Explanation. A Custom CNN With Five Convolutional Blocks Is Trained On A Dataset Of More Than 5,000 RGB Soil Images Covering Seven Major Indian Soil Types — Alluvial, Arid, Black, Laterite, Mountain, Red, And Yellow — Resized To 128×128 Pixels And Expanded Through Data Augmentation. On A Held-out Test Set Of 1,020 Images The Trained Model Attains An Overall Accuracy Of 92.45%, A Weighted F1- Score Of 0.9231, And A Macro F1-score Of 0.8351, With Per-class Accuracy Reaching 100.00% For Red Soil, 99.55% For Black Soil, And 95.09% For Yellow Soil. Together With The Predicted Class The Interface Reports A Confidence Value, The Full Probability Distribution Over All Seven Classes, And The Crops And Regions Associated With The Identified Soil, So That The Output Is Directly Actionable For Crop Selection And Land Management. The Prototype Was Validated Through Functional Test Cases Covering Dataset Loading, Preprocessing, Model Training, Prediction, And Visualisation, All Of Which Passed. Overall, The System Demonstrates That A Compact CNN Trained On Ordinary Camera Images Can Deliver Fast, Low-cost, And Reasonably Accurate Soil Identification Suitable For Field Use By Farmers And Extension Workers.

Published:

06-10-2026

Issue:

Vol. 26 No. 10 (2026)

Page Nos:

18 - 31

Section:

Articles

License:

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

How to Cite