Digital Image based Soil Organic Matter and Moisture Content Characterization for Precision Agriculture
Information on spatio-temporal variation of soil organic matter (SOM) and soil moisture content (SMC) is critical for applications in agriculture, forestry, land degradation management, environment protection, and land use planning. However, traditional laboratory characterization and measurement of these properties is expensive and time and labour-intensive. With developments of technology and computational power, digital image processing-based soil characterization has shown potential as an easy, fast, and inexpensive technique as colour and/or reflectance of soil can be attributed to numerous properties of soil including SMC, SOM, parent material, mineralogy, and texture. This thesis investigated various aspects of digital image processing for soil characterization in laboratory conditions including optimizing image features, and calibrating, validating and comparing various supervised regression and machine learning models for developing predictive relationships between image features and laboratory-measured SOM and SMC. The research showed strong potential of image-based soil characterization for applications in various fields including precision agriculture.