Submitted to: Biennial Workshop on Aerial Photography, Videography, and High Resolution Digital Imagery for Resource Assessment Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 11/2/2005
Publication Date: 2/1/2006
Citation: Wallet, B., Vogt, J.T. Graphical exploration of features to detect imported fire ant (Solenopsis spp., Hymenoptera: Formicidae) mounds in high resolution aerial imagery. Biennial Workshop on Aerial Photography, Videography, and High Resolution Digital Imagery for Resource Assessment Proceedings. Weslaco, TX, October 4-5, 2006.
Interpretive Summary: Airborne digital imagery is a potential tool for detecting imported fire ant mounds over large areas. Researchers in the USDA, ARS Biological Control of Pests Research Unit, in collaboration with Cooperative Research and Development Agreement partner Automated Decisions, LLC, have discovered key differences between fire ant mounds and other objects (or “false alarms) on the ground that will be useful for automated or computerized classification of mounds in aerial images. This research has the potential to greatly reduce the time needed to detect fire ant mounds in aerial imagery, allowing research and regulatory personnel to map fire ant infestations over large areas with minimal effort.
Technical Abstract: Automated Decisions, LLC is partnered with the USDA, ARS Biological Control of Pests Research Unit under Cooperative Research and Development Agreement to develop techniques for identifying and quantifying imported fire ant mounds in remotely sensed data. The problem of detecting of fire ant mounds using high resolution aerial imagery is complicated by the fact that indicative features vary widely with time, season, and background environment. In approaching this problem, we have chosen to use adaptive techniques that use different classifiers in different situations. This necessitates methods that allow for the rapid construction of classifiers based upon a broad set of spectral and shape based features. This paper discusses our use of Exploratory Data Analysis techniques to graphically construct classifiers. Our focus is upon the use of visualization techniques to identify features and projections of features. We present graphical results of our work as well as discuss advanced visualization techniques that we are currently implementing.