The Research Of Pesticide Residues On Hami Melon Surface Based On Hyperspectral Imaging Technology

Hami melon is one of the famous featured fruit in Xinjiang. Because of its unique fragrance, highly nutritious and superior antioxidant capacity of its seeds, it is therefore popular for the majority of consumers. Recently, food safety issues are the main topic people concerned about. This paper focuses on the research object Hami melon to study the pesticide residues, which is one of the food safety issues. Nowadays, the detection method of pesticide residues emerges in an endless stream, for example, Gas Chromntography, Liquid Chromntography, enzyme inhibition and so on, but these methods need to improve in detect time, cost and many more. The rapid, non-destructive and precision testing method is the ultimate goal. Thus, hyperspectral imaging techniques are employed to detect the pesticide residues on the surface of Hami melon, thereby providing theoretical bases for producing and manufacturing a rapid and nondestructive pesticide residue detector. The main contents are as follows:(1)The distilled water were made up pesticide sample of different types and concentrations pesticides. Pesticide samples were dropped on the surface of Hami melon and marked; the samples information were collected by hyperspectral imaging system at different light. To remove the absolute noises of the spectra, only the data of spectral range 450-1000 nm was used for analysis and extracted ROI region spectral data.(2) According to the extracted ROI region spectral data. 7 different types of pesticides under different lighting conditions Hami melon were sorted by using distance discriminant method, Bayes discrimination method, support vector classification and machine learning machine classification limit four kinds of prediction models and a 99.29% of identification accuracy was achieved.(3)The Hami melon with same pesticides and different concentrations under different lighting conditions were discriminated. Phoxim and Fenvalerate pesticides spectral data were extracted in Halogen and ultraviolet light sources. Then according the reflectance values of wavelength to select the characteristic wavelength. Same pesticides and different concentrations Hami melon under different lighting conditions were sorted by using distance discriminant method, Bayes discrimination method, support vector classification and machine learning machine classification limit four kinds of prediction models. The results of the distance discriminant method achieved identification accuracy of 100% for two kinds of pesticides. In non-linear discriminant methods, the accuracy of support vector machine higher than limit machine learning to distinguish Fenvalerate and Phoxim pesticide category, and the accuracy of support vector machine Phoxim pesticide concentration determination up to 100%. The results of the support vector machined method achieved identification accuracy of 86.67% for distinguish the concentration of pesticide Fenvalerate.(4) The impact of different light sources on the discrimination result was explored. This paper utilized four methods to detect pesticide residues on the surface of Hami melon. From the result, discriminant accuracy under the UV light is significantly higher than the discriminant accuracy under the Halogen light. The concentration sort of Fenvalerate residues was detected suitable under the Halogen light. And the concentration sort of Phoxim residues was detected suitable under the UV light.