Method for predicting contents of soluble solids of korla fragrant pears based on CARS-MIV-SVR

Abstract

The invention discloses a method for predicting the contents of soluble solids of korla fragrant pears based on CARS-MIV-SVR. The method for predicting the contents of the soluble solids of the korlafragrant pears based on the CARS-MIV-SVR comprises the steps that original hyperspectral images of all fragrant pear samples are obtained and subjected to black and white correction; areas of interestare extracted, and hyperspectral data are obtained; the contents of the soluble solids of all the fragrant pear samples are measured, and sample sets are divided; a standard normal variate method isused for preprocessing the data, and then the characteristic wavelength is extracted; by taking full-wavelength spectrum information and spectral information obtained through characteristic wavelengthselecting methods as input vectors, a support vector regression prediction model is established; and according to the model prediction result, model performance is evaluated, and the characteristic wavelength selecting method with the best prediction effect is selected. The method has the beneficial effect that the characteristic wavelength is screened through a combined algorithm of a competitive adaptive reweighting algorithm and an average influence value algorithm for modeling analysis.

Description

technical field [0001] The invention relates to the field of predicting the soluble solid content of Korla fragrant pear, in particular to a method for predicting the soluble solid content of Korla fragrant pear based on CARS-MIV-SVR. Background technique [0002] Hyperspectral imaging technology combines spectral information and image information, which can reflect the chemical composition and microstructure of samples. Many scholars at home and abroad have studied the non-destructive testing of fruits based on the spectral reflectance of samples, combined with data transmission interactive equipment and multivariate statistical tools. The research shows that the use of hyperspectral imaging technology can effectively detect common defects and quality indicators (such as hardness, , moisture, etc.) and maturity, etc. [0003] The traditional technology has the following technical problems: [0004] Throughout the existing research on the detection of soluble solids in Kor...

AI Technical Summary

Problems solved by technology

[0004] Throughout the existing research on the detection of soluble solids in Korla fragrant pears, some scholars use the Competitive Adaptive Reweighting Algorithm (CARS) to extract the characteristic wavelengths. In the CARS method, the variable regression coefficient will change due to the random selection of modeling samples. , the absolute value of the regression coefficient cannot fully reflect the importance of variables, thus affecting the accuracy of model detection

Images