Method for establishing prediction model of pit mud total acid based on hyperspectral image technology

Abstract

A method for establishing a prediction model of total acid in pit mud based on hyperspectral image technology. Firstly, hyperspectral image information in the near-infrared and visible light bands is collected, and the average spectral information of the region of interest (ROI) is extracted by ENVI software. Eliminate the error caused by uneven illumination, estimate the average spectral reflectance of each pixel in the image, and establish quantitative prediction models based on full-band and characteristic spectra respectively after no processing, SNV, and different modeling methods, and compare and analyze Each fitting model training set and test set determine the size of the coefficient and the root mean square error, thereby screening out the best model for the quantitative prediction of pit mud total acid. Monitoring provides a strong technical support.

Description

technical field [0001] The invention relates to a method for establishing a pit mud total acid prediction model based on hyperspectral image technology, and belongs to the technical field of solid-state fermentation index detection. Background technique [0002] Solid-state fermentation technology has a long history in my country, and it is also a unique brewing process in my country. In the liquor brewing industry, pit mud is the basis of fermentation, in which caproic acid bacteria are the main beneficial microbial flora, which use the nutrients in the fermented grains to produce acid and ester, and ethyl caproate is the main body of Luzhou-flavor liquor Therefore, pit mud plays a decisive role in the quality of Luzhou-flavor liquor in the production process, and it is more important to evaluate the quality of pit mud. Without good pit mud, it is impossible to produce superior quality wine. [0003] At present, many distilleries are faced with the situation of high acidi...

AI Technical Summary

Problems solved by technology

[0004] The deficiency of the patent: the patent detects the pH value, that is, the concentration of free hydrogen ions, so the pH is closely related to the type and content of acidic substances, and the characteristic absorption performance of the fermentation substrate to the spectrum is also reflected by the coloration carried by various acidic substances. Group performance, because the optical properties of a substance are generally expressed through functional groups, and different functional groups have different absorption strengths for different wavelength bands, which are related to the molecular structure

According to the pre-experiment and data query, the total acid shows its optical properties mainly through the carboxyl group, and has strong absorption in the near-infrared band, while the patent only collects images and spectral information in the visible light region; no full spectrum Compared with the characteristic spectrum, the spectral information corresponding to the characteristic wavelength is directly selected as the input variable of the model, which is easy to miss important information; without preprocessing and denoising, it is easy to include non-sample signals into the calculation; the modeling method is also relatively simple

Images