A large-scale agricultural concept with small-scale mechanization requires an optimal monitoring of the fields, which goes beyond current methods and technology. In a large-scale agricultural system, the observation could pose a bottleneck. The good observation of weeds and diseases is a labour-intensive process for a farmer.
Within the project an innovative image recognition concept is used. The concept is designed by a PhD student of WU – Farm technology and based on machine learning technology. The concept has been successfully tested in a small scale environment and shows promising results. In the SMARAGD project this technology is used as the basis for a monitoring system for:
- the recognition of weeds and volunteer potato in sugar beets
- the recognition of early stages of crop diseases. The first focus for crop diseases is on the recognition of phytophthora in potatoes.
The monitoring itself can be done by small autonomous vehicles, as well as by drones. The Work package modelling (WP 1) should provide insight in the required capacity. The frequency of monitoring is determined by the need for timely discovery. The translation from the monitoring into actions is foreseen in two ways: First option is real-time control of weeds on the basis of their recognition and second is the translation to a task where field information leads to a site-specific treatment at a later time. For weeds, a risk analysis takes place based on the consequences of not or delaying weed combating (and the optimal site-specific dosage and timing of application). For diseases and pests, the contamination risks (expansion of the hearths) and the specific use of preventive or curative means (including the timing thereof) are examined.