Estimating Organic Matter, Soil Surface Roughness, Plant Stand Count, and Inter-plant Spacing from High-Resolution RGB Imagery

Precision agriculture is adopting site-specific agriculture practices reducing farm inputs and increasing productivity. Site-specific agriculture requires information about every inch of the field. Provided the large farm holding of the Canadian Prairies, getting extensive information about the field is an uphill task. Currently, limited samples are taken from the field resulting in low spatial resolution. Thanks to progress in economical image acquisition methods, acquiring high-resolution field imagery has become more accessible. The project aims to use high-resolution field imagery to reduce the burden of manual soil sampling and field scouting. This project studies relationships of different field attributes and investigates whether soil attributes like surface roughness and organic matter content can be predicted for RGB field imagery using machine learning and deep learning methods.
Similarly, plant stand count and inter-row plant spacing are estimated from RGB imagery. Additionally, the project studies seed mortality as a function of plant stand count, inter-plant spacing and soil attributes. Lastly, yield is estimated at the early growth stage of plants through non-destructive biomass estimation using high-resolution RGB ground imagery.

Faculty Supervisor:

Abdul Bais

Student:

Partner:

Croptimistic Technology Inc

Discipline:

Engineering

Sector:

Technology; Public Service, Policy, and Governance; Agriculture and Food

University:

University of Regina

Program: