Abstract: 【Aim】 To enhance the recognition effect of Bactrocera dorsalis Hendel (Diptera, Tephritidae) on the background of fruits, the difference in the reflectivity spectra of the insect and different fruits was studied. 【Methods】 The reflectivity spectra of B. dorsalis and 16 species of fruits were first measured in the wavelength range of 400-2 500 nm using the UV-Vis-NIR spectrometer. Then images of the flies on the fruits were taken respectively under the illumination of narrow-band light with the central wavelengths of 565 and 827 nm as well as sunlight, and binarization of the images was made according to Otsu algorithm. 【Results】 We found that the reflectivity of B. dorsalis increased slowly with the wavelength, and approached to the maximum (<40%), while the strongest reflectivity peak of the sixteen species of fruits was located between 777 and 896 nm entirely or partially. The average maximum reflectivity of each species of fruit ranged from 41.10% to 97.89%, while that of B. dorsalis was approximately 30%, leading to a higher difference in this NIR (near-infrared light) range. It was found that B. dorsalis was dark on the large-area bright background of fruits in the images taken under the illustration of narrow-band light with a central wavelength of 827 nm, which can be recognized easily because of a high contrast between B. dorsalis and the fruits. Contrarily, under the illustration of sunlight and narrow-band light with a central wavelength of 565 nm, more black spots were found in the images of the fruits which were confused with the dark images of B. dorsalis, or the insect appeared as a white spot, causing recognition failure. 【Conclusion】 We confirmed that the image contrast between B. dorsalis and fruits, and the segmentation effect of B. dorsalis can be enhanced under the illumination of the NIR narrow-band light.

Key words:  Bactrocera dorsalis, fruit, reflectance spectroscopy, near-infrared light (NIR), image identification