2020
sUAS Weed Mapping in Soybeans
Home
Contributor/Checkoff:
Ohio Soybean Council
Category:
Sustainable Production
Keywords:
AgricultureCrop protectionHerbicide
Parent Project:
This is the first year of this project.
Lead Principal Investigator:
Scott A. Shearer, The Ohio State University
Co-Principal Investigators:
Project Code:
20-R-11
Contributing Organization (Checkoff):
Ohio Soybean Council
$50,000
Institution Funded:
The Ohio State University
$50,000
Brief Project Summary:

This project helps growers identify and map herbicide-resistant weed escapes using small UAS as a precursor for targeted eradication. Weed escapes simply refers to weeds that survive weed management practices. Most weed species produce prolific seed. It’s easier to control weed escapes before they build the soil seed bank. The effort includes building a reference library of herbicide-resistant weed escapes that occur in Ohio soybeans, training Convolutional Neural Nets for sematic segmentation of NADIR imagery generated from fixed wing sUAS overflights, using this technology to map weed escapes and developing methodology for real-time classification of images on-board of the sUAS. The project also includes field tests.

Key Benefactors:
farmers, agronomists, extension agents

Information And Results
Project Deliverables

We will construct a database of weed images. A database is the first key element in any Artificial Intelligence (AI) application, which will be used for training CNNs. Given the increased use of AI applications in precision agriculture, development of a database is a key step for continued use of AI tools. The image library of weed escapes collected in Ohio will support weed classification.

2. A semantic segmentation algorithm will be developed to classify pixels in the images of weed captured by sUAS. Semantic segmentation algorithm developed in this project will be used to identify weed location. In addition, the same algorithm can be used in other applications such as land use or crop health assessments.
5 | s UAS Mapping of Weed Escapes


3. An algorithm will be used to map of weed escapes. This map-based approach ensures adoption of sustainable weed management practices to control herbicide resistant weed escapes.

4. The architecture of the trained CNN algorithm developed in this project to classify weeds has flexibility to expand and be re-trained to include additional images of different weed classes. The CNN algorithm developed is scalable which in turn supports commercialization.

5. The proposed sUAS system will include the capacity for real-time on-board image classification. Currently, the requirement to use secured digital (SD) cards and other memory devices makes transfer and processing of remote sensed imagery cumbersome. But, real-time classification makes these imagery data more valuable given the immediacy of processing.

6. The proposed systematic approach for weed mapping and identification is based on several assumptions required to maximize effectiveness, accuracy, and reliability. The system is scalable for incorporation in to commercial farming operations. With the end goal of field testing and evaluation, the assumptions will be validated in support of intellectual Property (IP) generati

Final Project Results

The United Soybean Research Retention policy will display final reports with the project once completed but working files will be purged after three years. And financial information after seven years. All pertinent information is in the final report or if you want more information, please contact the project lead at your state soybean organization or principal investigator listed on the project.