A software architecture towards automated data-driven multi-resolution crop scouting with an unmanned aerial system

Crop scouting and field monitoring is necessary to track the status of crops planted in the field throughout the season. Developing a sampling plan to determine the path through the field and the number of points to sample helps minimize the amount of time spent scouting while maximizing the quality of information collected. However, these practices do not sufficiently account for unexpected costly events such as crop damage from extreme weather. By using unmanned aerial systems (UAS) to obtain aerial imagery of the field, agronomists can hedge against unexpected events and develop scouting patterns driven by contemporary data. Yet, widespread adoption of UAS technology in precision agriculture is impeded by the lack of knowledge to interpret data for agricultural decision making and the complexity of operating UAS. In order to realize the use of UAS in practical application, it is necessary to employ an autonomous UAS for crop scouting that optimizes turn-around time and quality of information obtained. This thesis proposes a software architecture for UAS flight planning that can provide a two-stage flight mission consisting of a high altitude scout flight over a large area followed by a low altitude inspection flight at a limited number of places of interest deemed high priority according image analyses obtained in the scout flight. This would allow a large area to be covered in a short amount of time while also providing imagery with finer ground resolution for more accurate interpretation. An experimental implementation with digital imagery was developed as a proof of concept of this software architecture using well-established algorithms