A rapid recognition method of Auricularia auricula varieties based on near-infrared spectral characteristics

Edible fungus is one of the significant foods for human beings. The rapid recognition of fruit body phenotypic characteristics is of great significance, theoretical and practical value for the classification of edible fungus, evaluation of germplasm resources, breeding, and intelligent cultivation. The existing morphological-physiological phenotype detection for edible fungus' growth process is not systematic enough and lack of rapid and accurate detection methods for the fruit body varieties of Auricularia auricula. At present, the operation process of DNA molecular marker technology is complex, time-consuming, laborious, and expensive, moreover, the traditional artificial judgment of fruit body varieties based on naked-eye observation and experience easily led to errors and inaccuracy. Therefore, a rapid recognition method of Auricularia auricula varieties was proposed based on the characteristics of near-infrared spectroscopy. First, the four Auricularia auricula varieties of Cheng D, Hei Feng, Hei Shan, and Xu 1 were taken as the research objects, the near-infrared spectral data of Auricularia auricula were scanned by Fourier transform near-infrared spectrometer (Tango). Second, the detrend(DT)method was used to pre-process the raw spectral data. Then the competitive adaptive reweighted sampling (CARS) algorithm was applied to extract 131 effective characteristic wavenumbers from the pre-processed spectral data. Finally, the radial basis function (RBF) neural network of type 131-12-4 was constructed to recognize Auricularia auricula varieties. The research results showed that the accuracy of DT-CARS-RBF model for recognizing Auricularia auricula varieties was 99.32%, the root mean square error value was 0.0908, and the running time was 0.000956 s. This achievement established a rapid, efficient, and accurate method for the recognition of Auricularia auricula varieties, which provided a theoretical basis and technical support for the rapid of Auricularia auricula varieties for the classification of edible fungus and evaluation of germplasm resources, breeding, and cultivation.