Location: Crop Production Systems Research
Project Number: 6066-30500-003-004-S
Project Type: Non-Assistance Cooperative Agreement
Start Date: Jun 26, 2023
End Date: Jun 25, 2028
Objective:
Unmanned Aerial Vehicles (UAV), a.k.a. drones, have been used for monitoring large agriculture fields to obtain images of target plant species and spray herbicides as required. However, cleaning up the obtained images and extracting features must be done using a central station that limits such monitoring frameworks from real-time sensing in spatiotemporal applications.
Objective: To develop a low-cost Application-Specific Integrated Circuit (ASIC), Agri-eye, integrate it with unmanned aerial vehicles approved for federal government use, test the system for agricultural applications, including differentiating between common cover crops grown in Mississippi, distinguishing between row crops, and separating row crops from broadleaf weeds and grasses. The evaluation metrics will be based on how well the system identifies plants with different crop morphologies.
New Objective: Added funds will be used to develop and deploy an end to end, AI enabled precision weed management system that includes an advanced model repository for weed and cover crop detection, a stakeholder facing web and mobile platform for decision support, and a closed loop UAV based precision spray system that translates real time analytics into targeted, site specific herbicide applications.
Approach:
Feature Identification of Crops and Selected Broadleaf and Grass Weeds (Year-1): As the first step in this research, we will evaluate two types of drones that can be integrated with low cost cameras to run embedded machine learning algorithms. PI from UT Tyler will work with USDA ARS to evaluate the Blue UAS-cleared drones, such as the IF750, and lighter drones, like RYZE Tello Edu Drones. These drones will collect images from the field using the camera on the drones and integrate the lowcost cameras and sensors as attachments/fixtures to help evaluate trained models for real-time surveillance. We will identify unique morphological and environmental features such as temperature, humidity, weather, color, and dimensions (height,width, and size) for each crop and select weeds in the targeted field(s) with the help of imagery acquired by the drones. Feature identification will be achieved through year-round monitoring and collection of images, along with integrated sensors, which can help automate identifying the crop type and weeds.
Developing the agri-eye for the drones (Year-2): This research aims to develop a workable camera chip with an on-device learning capacity for differentiating in-field plant species using metadata collected from the integrated sensors. The proposed ASIC will contain the sensors required to identify the features for sensing windows. The development phase will
include the ASIC design along with a user interface, such as a mobile app/ web interface that can function as a middleman between the existing drone interfaces and the proposed ASIC. This will help in developing a sustainable ecosystem for the proposed chip to be integrated in any framework.
Adding funds for the current agreement will accelerate research that lowers herbicide use by expanding mechanization and automation. This work directly advances a major USDA research priority by supporting farmers and ranchers to provide safe, nutritious, and secure food supply. This approach begins with harmonizing multisensor UAV datasets and retraining high accuracy detection models validated through standardized performance metrics and real time inference benchmarking. These analytics feed directly into an API driven web and mobile platform that supports secure data ingestion, geospatial processing, prescription map generation, and offline capable field use, refined through iterative stakeholder centered usability testing. The resulting prescriptions drive a closed loop precision spray drone system that links detection, mapping, and autonomous actuation, with field validation quantifying accuracy, operational efficiency, and post treatment weed suppression. The additional funding received will enable the postdoctoral researcher at the University of Texas at Tyler to continue advancing this work.