Development of a Plant Monitoring System Using a Drone-Equipped Agricultural Robot*

This study proposes a plant monitoring system to automate agricultural works by robots in plant factories. This system uses a drone-equipped agricultural robot to monitor plants. This robot has a camera-equipped drone tethered with a power supply cable. Therefore, the drone can fly without worrying about its battery capacity. As a case study, this system targets a strawberry greenhouse. YOLO (You Look Only Once) v4 is used to detect strawberry plants. Created detection models were evaluated by four-fold cross-validation, the average accuracies of flowers, immature and mature fruits were 0.826, 0.841 and 0.920, respectively. As the performance of this system, the numbers of strawberry plants were counted based on the detection results, and the counting results were evaluated by comparing with the actual numbers. The accuracies of flowers, immature and mature fruits were 0.963, 0.961, and 0.972 respectively. Even if strawberry plants cannot be detected from one viewpoint, if they can be detected from another viewpoint, the task of detecting them is achieved. The results of this experiment demonstrated the effectiveness of a plant monitoring system using a drone-equipped agricultural robot.