Soil Classification through Advanced Image Processing and Deep Learning Models

Dialani, Sanjay Girish (2023) Soil Classification through Advanced Image Processing and Deep Learning Models. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (6MB) | Preview
Preview	PDF (Configuration Manual) Download (758kB) | Preview

Abstract

The research study aims to design a deep learning model, addressing the need for accurate and efficient methods to classify various types of soil. The research work utilizes advanced image processing techniques along with deep learning models, integrating two diverse datasets encompassing Alluvial Soil, Black Soil, Cinder Soil, Clay Soil, Laterite Soil, Peat Soil, and Red Soil.

Motivated by the significance of understanding soil types for agricultural planning and environmental management, this research aimed to develop robust models capable of accurately classifying soil types based on image analysis. Three different models were devised and implemented. The first model employed Convolutional Neural Networks (CNN) to provide initial intuition into the classification task. The second model utilized Transfer Learning, specifically leveraging the VGG16 architecture which is pre-trained on ImageNet weights. Whereas, the proposed third model (SoilNet model) introduced a unique approach involving pre-processing techniques such as histogram equalization, Gaussian blur, and median filtering before applying Transfer Learning.

The results illustrate the efficacy of the models in capturing complex patterns within soil images. The CNN model achieved a validation accuracy of 88.46%, while the Transfer Learning model with VGG16 exhibited enhanced performance with an accuracy of 93.85%. Notably, the third model, incorporating pre-processing steps, outperformed the others, achieving an impressive accuracy of 95.38%.

The research work aligns with the current state of the art by showcasing the applicability of deep learning techniques in soil classification and building the advanced technique to address the gaps in the previous works. In practice, the key benefit lies in the development of accurate models that can assist in efficient soil type identification, helping in agricultural decision-making and environmental planning. However, some aspects, such as the impact of varying environmental conditions on model performance, present opportunities for future research in the field of soil classification.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Jilani, Musfira UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science S Agriculture > S Agriculture (General) H Social Sciences > HD Industries. Land use. Labor > Specific Industries > Agriculture Industry G Geography. Anthropology. Recreation > GE Environmental Sciences > Environment Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	07 May 2025 14:23
Last Modified:	07 May 2025 14:23
URI:	https://norma.ncirl.ie/id/eprint/7508

Actions (login required)

View Item