An Automatic On-site Fire Ant Screening System Sanqiang Zhao1,2, Yongsheng Gao2,1 Terry Caelli3, Fabian Bracco4 1Queensland Research Laboratory, NICTA, Australia 3Victoria Research Laboratory, NICTA, Australia 2Institute for Integrated and Intelligent Systems 4Griffith School of Engineering, Griffith University, Griffith University, Brisbane, Australia Brisbane, Australia {s.zhao, yongsheng.gao}@griffith.edu.au
[email protected] Abstract—This paper proposes the first attempt for semi- established in an area and leading ultimately to environmental automatic screening and identification of red imported fire ants and industrial devastation. Moreover, human expertise that is (Solenopsis invicta) in Australia. As an exotic ant species to required to correctly identify fire ants is very limited. For Australia, fire ants were imported from South America in 2001 example, there are only around four full-time entomologists and have since been regarded as dangerous pests that could devoted to fire ant identification at Biosecurity Queensland severely damage the environment and many industries. We Control Centre (BQCC). Therefore, constant monitoring, followed two of the three major identification keys defined by prompt detection and automatic identification of fire ants in entomologists and proposed: 1) A fusion of two different image situ play important roles for the control and eradication of fire features (i.e., the perpendicular median intensity and the ants across Australia. perpendicular width) for antenna segment detection; and 2) A weighted histogramming of micropattern features for petiole For computerized insect identification, there are two broad classification. Our experimental results show that automatic on- categories, i.e., hierarchical top-down classifications using site fire ant screening is feasible and the proposed weighted taxonomic keys, and case-by-case bottom-up searching using histogramming of micropattern features performs better than the known species samples.