In this paper the wing vein pattern recognition of Rhagoletis pomonella species complex was investigated. There are five taxa studied: R. mendax, R. zephyria, R. cornivora and apple race and hawthorn race of R. pomonella, The right wing was removed and mounted on a slide, and then its vein length was measured. Twenty two landmarks were labeled and thirty four measurements were made on the right wing of the tephritid. Bayesian decision theory is a fundamental technique for pattern recognition and classification. The Bayes classification models were built to solve our problem of wing pattern recognition based on the collected vein dataset. Two methods were used to estimate the joint probability density functions for each taxon: parametric and Parzen window-based approaches. The Parzen’s model was implemented in a neural network called Probabilistic Neural Network. The results show that parametric method has a better performance than Parzen’s. The average success rate of the classification system is 73.1% with performance confidence interval [60.7%, 82.5%]. Among five taxa R. cornivora and R. mendax can be identified without error (100% success rate), and apple and hawthorn races are mixed up and cannot be separated from each other.
Species 1: Diptera Tephritidae Rhagoletis pomonella (apple maggot)
Keywords: wing, pattern recognition
The ESA 2001 Annual Meeting - 2001: An Entomological Odyssey of ESA