Skip to main content
ARS Home » Southeast Area » Stoneville, Mississippi » Cotton Ginning Research » Research » Publications at this Location » Publication #388889

Research Project: Development and Evaluation of Novel Technologies to Improve Fiber Quality and Increase Profitability in Cotton Processing

Location: Cotton Ginning Research

Title: Developing a UAV sourced database for AI trash detection in cotton fields

item Blake, Cody
item Donohoe, Sean
item Thomas, Joseph

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 11/4/2021
Publication Date: 1/5/2022
Citation: Blake, C.D., Donohoe, S.P., Thomas, J.W. 2022. Developing a UAV sourced database for AI trash detection in cotton fields. Meeting Abstract. 2022 Beltwide Cotton Conferences, San Antonio, Texas P.217.

Interpretive Summary:

Technical Abstract: The US cotton industry has seen an increase of extraneous matter such as plastics, twine, and other trash. In turn, researchers have focused efforts on detecting and removing that trash in different areas of the cotton harvesting and ginning processes. The objective of this work is to develop a database of UAV sourced RGB images to train an AI trash detection system. The images collected need to be preprocessed for size, shape, and format. Images containing trash also need to be annotated, which is a time-consuming process. GPS location embedded within the image is useful in the determination of portion of the field where the trash needs to be collected. The optimization of this process is labor intensive as thousands of images need to be collected and processed before training of the AI can occur. After training, the AI system requires assessment on several performance metrics and tuning as needed.