- Decision support systems (DSS) for agriculture 4.0
- Smart agricultural machinery and low-cost precision agriculture
Deep Learning and computer vision are now being applied to digital farming issues such as weed recognition, detailed prediction of water requirements and nitrogen fertilisation, and estimating evapotranspiration. Together with low-cost wireless sensor networks, AI has provided encouraging results such as the recognition of vine diseases based on the analysis of photographs of foliage; measurements of temperature, leaf wetness and humidity have made possible the early detection of diseases such as black rot; and the remote control of paddy field crops. Pest and weed monitoring can be carried out using spectroscopy techniques in the field, both for plants and soil. Working with selected end users, we are defining best practices for the creation of image and measurement databases with considerable information content, highly specific to the application concerned and of appropriate size. Our approach involves a training phase based on high-performance hardware, followed by inference carried out as far as possible on edge computers, also in real time, using NPU (Neural Processing Unit) accelerators and other specialised AI devices.
The DSS will first be applied and tested at an experimental farm and, once an appropriate level of development has been reached, tested at several pilot farms, with a view to disseminating new developments.