Soil Analysis via Remote Sensing and Artificial Intelligence for Precision Regenerative Agriculture

Abstract

<p>Soil electrical conductivity maps were generated for greenspace in the Greater Toronto Area using a conditional generative adversarial network (CGAN), which is a form of deep learning where one neural network is used to train another. The results of the analysis show that the model can accurately predict soil conductivity 34.6% of the time. It could possibly be strengthened with the inclusion of more electromagnetic bands in the supervised classifications used to train the network, such as the infrared spectrum, as well as Light Detection and Ranging (LiDAR) data. This three-dimensional imagery should be considered, as the model is not optimized when soil is obscured by foliage. Generally, these two data types are commercially available and commonly used for the analysis of greenspace. Microdrones can potentially be equipped with computer vision-enabled sensors operating this neural model to iteratively analyze soil types and complete aerial cropping.</p>