Machine learning to determine soil properties: a smartphone-based solution for stakeholders

Soil properties (such as organic carbon) can be estimated using modern data mining and modelling approaches. The proposed approach correlates soil colour and image structure metrics extracted from smartphone and tablet camera images, as well as environmental factors. This principle was used as a foundation to develop a mobile phone app which measures Scottish soil organic carbon content. This project aims to improve on this concept with a model that enhances the performance and functionality, and integrates it with the system hosting the app. Furthermore, the project proposes analysing the pH, bulk density, particle size distribution and permeability of the soil utilising images. Upon the development of the model, a new SolEst app will be developed which provides an environmentally-friendly and costless platform for land managers working in the construction and agriculture sectors to analyse these and other soil properties using their mobile phone. This interdisciplinary project merges geotechnical engineering, digital technology and artificial intelligence to introduce a cutting-edge method for soil analysis using smartphone which is widely available.

The project is the combination of modelling, laboratory and experimental work which enables the PhD candidate to acquire skills of programming, working in-situ and in laboratory during the course of the program.