Enhancing agricultural object recognition in UAV multimedia with integrated sensing path planning approaches.

Precision Agriculture reduces the amount of inputs applied to the crop and hence increases the benefit obtained by the farmer while lowering the carbon footprint of the campaign and producing more sustainable food. UAVs have gained importance in the last years because of the low cost and field-specific solutions they provide in agriculture. Nevertheless, most of the machine learning methods applied until the date to UAV data require a lot of human intervention, which is expensive, time-consuming, non-objective, and prone to errors. Furthermore, the current UAV missions are planned without considering the optimal coverage trajectory and the final purpose of the data collection, which is crucial to acquire high-quality data using the least battery and flight time. Hence, there is a lack of knowledge on planning optimal UAV flight paths for specific data-acquisition purposes within PA contexts. During this research, we aim to plan efficient flight paths to improve the quality of the input multimedia which will be used for object detection purposes mostly focusing on woody crops. This Ph.D. project will focus on designing an efficient UAV path planner algorithm to enhance the input multimedia that is later used for object detection purposes in woody crops. The research line will target the application of object detection techniques to a standard multimedia input acquired following the UAV-provider path planner and compare the detection metrics with the multimedia input collected using the efficient path planner algorithm designed during this Ph.D. We aim to improve the quality of the input multimedia to enhance object detection.