Skip to main content
ARS Home » Research » Publications at this Location » Publication #114010
Title: DETECTION OF COCKLEBURS (XANTHIUM STRUMARIUM L.) IN SOYBEANS USING HYPERSPECTRAL IMAGERY

Author
item VARNER, B - ITD, URBANA, IL
item GRESS, T - ITD, URBANA, IL
item COPENHAVER, K - ITD, URBANA, IL
item Wax, Loyd
item SPRAGUE, CHRISTY - UNIV OF ILLINOIS
item TRANEL, PATRICK - UNIV OF ILLINOIS

Submitted to: International Conference on Precision Agriculture Abstracts & Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 8/20/2000
Publication Date: N/A
Citation: N/A

Interpretive Summary: Weed detection and management are major concerns for soybean growers throughout the United States. Remote sensing could potentially make weed detection and management more efficient. We conducted a study on an experimental farm to attempt to detect cocklebur (Xanthium strumarium) in soybean using hyperspectral imagery. Cockleburs were purposely grown with the soybean to guarantee their presence and the study are was also used as a soybean/cocklebur competition study. Detection of the cockleburs and soybeans was accomplished by collecting visible and near infrared spectral signatures from an airborne hyperspectral sensor. The data used in this study were taken with a real-time digital airborne camera system on two dates in July of 1999. These signatures were used to build a spectral signature database. Classification and other image analysis techniques were investigated to optimize extraction of information that indicates the presence of cockleburs with the soybean plots. Accuracy assessments were conducted on these classifications to show the effectiveness of these techniques. These findings add to the database for remote sensing of weeds in soybean field and provide new information that may allow for better mapping and control of troublesome weeds in soybean fields. The results should be useful to federal, state and industry scientists who are involved in developing improved weed management systems.

Technical Abstract: Weed detection and control are major concerns for soybean growers throughout the United States. Remote sensing could potentially make weed detection and control more efficient. A plot study on a 1.5-acre field at the University of Illinois Crop Sciences Research and Education Center was conducted to attempt to detect cockleburs (Xanthium strumarium L.) in soybeans using hyperspectral imagery. Cockleburs were purposely grown within the soybean to guarantee their presence. Weed detection for the cockleburs and soybeans was accomplished by collecting visible and near infrared spectral signatures from an airborne hyperspectral sensor. The data used in this study was taken with Spectral visions' real-time digital airborne camera system (RDACS/H-3) on July 11 and July 27, 1999. These signatures were used to build a spectral signature database. Classification and other image analysis techniques were investigated in ENVI 3.2 and ERDAS IMAGINE 8.3 to optimize extraction of information that indicates the presence of cockleburs within the soybean plots. Accuracy assessments were conducted on these classifications to demonstrate the effectiveness of these techniques.