FarmBeats: Automating Data Aggregation

Farm Policy Journal | , Vol 15: pp. 7-16

Digital agriculture offers one of the most promising approaches to address the challenge of sustainably increasing food production by 70% by 2050 (from 2010 production levels). Using the latest advances in artificial intelligence (AI) and machine learning (ML), the farmer can be empowered with predictions that can improve farm processes, from planning until harvest.
Satellite data and remote sensing techniques can provide agricultural insights, by using advanced image processing algorithms and AI algorithms on multiple spectral bands in satellite imagery to estimate crop health. However, satellite imagery alone is unable to capture all the data from the farms. Recent work has investigated the use of in-field sensors and imagery to complement satellite data, along with unmanned aerial vehicles (UAVs), cameras and sensors on tractors. These data are streamed to the cloud using the latest Internet of Things (IoT) technologies, where they are processed to provide valuable insights to the farmer.
However, there are two key challenges in enabling this IoT-enabled vision of data-driven farming. First is the ability to get data from the farm, as most farms have poor Internet connectivity. The second is how to make data from different modalities actionable by the farmers. The heterogeneous sensor streams need to be merged and analysed together with satellite data. In addition, data collection and analysis need to be done in a way that does not add to the farmer’s workload, but instead streamline efficiency.
The FarmBeats solution at Microsoft uses new technologies, such as TV white spaces and Azure IoT Edge, to collect large amounts of data from the farm via sensors, tractors, cameras, drones and other devices. FarmBeats then develops new AI & ML algorithms (trained on this data), along with any available remote sensing data, to provide unique, actionable insights to farmers which can improve productivity.