Research on landmark extraction technology in identification of fruit flies (Diptera: Tephritidae)

Application of mathematical morphology has speeded up into identification of species, including insects. In this study, we use TpsDig, a software product for feature extraction, to locate 17 landmarks with their coordinate value on digital images of wing samples from four species of fruit fly (Bactrocera dorsalis, B. cucurbitae, B. tau, B. scutellata) in Tephritidae. Twenty two geometrical morphometric parameters are evaluated to analyse the morphological divergence among the four species. Three standardized canonical discriminant functions are formulated: Y₁=-0.114X₁+1.121X₂+0.613X₃-0.067X₄-0.019X₅, Y₂=0.515X₁-0.001X₂+1.379X₃-2.242X₄-0.795X₅, and Y₃=0.962X₁+0.071X₂+0.589X₃+0.091X₄-0.780X₅. It proves that correctness of the samples in species identification can reach at 98.1%.