Autonomous scouting using unmanned aerial vehicle

Abstract

Implementations are described for improving accuracy and efficiency in anomaly detection during autonomous scouting of crop fields. A method is provided and includes: operating an unmanned aerial vehicle (UAV) to fly along an initial flight path above a field, where the UAV collect images for the field and an edge-computing device disposed on (or included in) the UAV processes the images using a coarse-grained machine learning (ML) model to determine whether target region(s) are detected within the field. A reversed flight path can be determined based on the detected target region(s), and the UAV can fly along the reversed flight path from a point corresponding to an ending point of the initial flight path, to collect target images particularly for the detected target region(s). The collected target images can be processed using a fine-grained ML model, to more fully characterize the detected target region(s).