Real-Time Plant Recognition and Crop Row Navigation for Autonomous Precision Agricultural Sprayer Robot

This paper presents a vision-based selective spraying technique for an autonomous agricultural sprayer robot. In traditional methods, excessive chemical spraying cause deleterious effects on human health, environment and becomes uneconomical. In order to reduce the agrochemical wastage encounters in a broadcast spraying, a selective plant spraying method is used for crop chemical treatment. The sensor-based approach assisted with YOLOv7 model is deployed on a custom designed robot for recognizing and localizing the lettuce plants in field. The PID-based pressure controller is designed that minimizes the undesirable fluctuations cause by the opening/closing of solenoid-valve-nozzles (SVNs) during spraying. Thus the nozzle's spraying quality is maintained by keeping the pressure constant to a desired value. A visual servoing scheme for row tracking is presented that uses the detected plant's spatial features. The robustness of the visual-based navigation is validated in the real field experiments.