Abstract

Small- and medium-scale specialty crop farms often lack affordable, effective options for inter- row weed control. Commercial herbicides are frequently unapproved for these crops, forcing growers to rely on costly manual labor. This study uses horseradish as a case study to develop a low-cost, modular precision sprayer suited to small- and medium-scale operations. The system costs $750 USD when integrated with the farm ng Amiga robotic platform, but it can also be adapted to other platforms, such as an ATV, a Gator, or a pumpkin harvest wagon. The system specifically targets inter-row weeding using a 25-gallon tank, a 2.4 GPM pump, and an electric solenoid valve actuated by an ESP32 microcontroller. For real-time weed identification, the sprayer is equipped with an NVIDIA Jetson Orin Nano and dual 1080p cameras. The vision system was tested using an Excess Green color-based segmentation index and a YOLOv8n machine learning model. Field testing in horseradish evaluated the sprayer's accuracy at practical speeds, achieving spraying accuracy of 97.6% at 1 mph, 80.2% at 2 mph, and 51.5% at 3 mph. The ExG algorithm processed faster than the YOLO model, as it is computationally less expensive and more practical for inter-row weeding, where everything growing within the row is considered a weed. The modular design allows the same core system to be reconfigured for other row and wide-spaced specialty crops, including pumpkin, without major redesign. Further development will focus on integrating the platform with ROS2, expanding the number of nozzles, testing across additional crops such as pumpkin, and adding more robust hardware for reliable field performance.

Keywords

Applied Precision Agriculture

DOI

10.5703/1288284318729

Share

COinS
 

Low-Cost Modular Precision Sprayer for Specialty Crop Weeding

Small- and medium-scale specialty crop farms often lack affordable, effective options for inter- row weed control. Commercial herbicides are frequently unapproved for these crops, forcing growers to rely on costly manual labor. This study uses horseradish as a case study to develop a low-cost, modular precision sprayer suited to small- and medium-scale operations. The system costs $750 USD when integrated with the farm ng Amiga robotic platform, but it can also be adapted to other platforms, such as an ATV, a Gator, or a pumpkin harvest wagon. The system specifically targets inter-row weeding using a 25-gallon tank, a 2.4 GPM pump, and an electric solenoid valve actuated by an ESP32 microcontroller. For real-time weed identification, the sprayer is equipped with an NVIDIA Jetson Orin Nano and dual 1080p cameras. The vision system was tested using an Excess Green color-based segmentation index and a YOLOv8n machine learning model. Field testing in horseradish evaluated the sprayer's accuracy at practical speeds, achieving spraying accuracy of 97.6% at 1 mph, 80.2% at 2 mph, and 51.5% at 3 mph. The ExG algorithm processed faster than the YOLO model, as it is computationally less expensive and more practical for inter-row weeding, where everything growing within the row is considered a weed. The modular design allows the same core system to be reconfigured for other row and wide-spaced specialty crops, including pumpkin, without major redesign. Further development will focus on integrating the platform with ROS2, expanding the number of nozzles, testing across additional crops such as pumpkin, and adding more robust hardware for reliable field performance.