Mašīnmācīšanās balstīta mērījumu novērtēšanas pieeja precīzajai lauksaimniecībai

The Thesis puts forward a novel methodology employing deep learning techniques to facilitate the estimation of diverse microclimate metrics through the utilisation of thermal radiation imagery, thereby diminishing reliance on direct sensor-based measurements. The research proposes an open-source prototype data acquisition system, which facilitates the automated acquisition of radiometric thermal images alongside timely synchronised measurements of air temperature, relative humidity, soil moisture, and illuminance. The research involved the training and comparative analysis of three distinct deep learning architectures: a convolutional neural network, a vision transformer, and a hybrid convolutional vision transformer.