Study On The Detection Method Of Soluble Solid Content In Red Fuji Apple

China’s apple production ranks in the forefront of the world,while apple export pattern in the international market is far lower than the status of the developed countries in Europe and America.The reason is that China’s apple industry has a low degree of commoditization and its commercial processing mode is backward in the meanwhile.The post-harvest processing technology needs to be improved in order to enhance the market competitiveness of China’s apple industry in the world.The quality of its products is the primary focus of post-harvest treatment,and soluble solids are the emphasis of the product quality system.Near-infrared spectrum detection technology,integrating the advantages of lower cost,accurate and stable detection result and low loss,is one of the most mainstream detection solutions in the world.In this paper,near infrared spectroscopy technology is applied as the experimental method to realize the prediction of SSC results.In this paper,the collected apple near infrared spectroscopy is used as the data sample.First,the principal component analysis-the Mahalanobis method(PCA-MD)is used to remove the abnormal samples in the data according to the threshold restriction;then the spectral noise is processed,where the signal is filtered using the Savitzky-Golay convolution smoothing method to remove the noise in the signal Removal;Finally,using multiple scattering correction(MSC)to perform baseline correction on the spectrum of apple samples.The pre-processed spectral information still contains a lot of invalid wavelength information.In this paper,the successive projection algorithm(SPA)and genetic algorithm(GA)are used to filter the wavelength for the purpose of improving the prediction accuracy and stability of the model,and the extraction results are evaluated using two indicators: correlation coefficient and root mean square error.It is concluded that GA screening results are more advantageous than SPA screening results.After screening the wavelengths,two algorithms,Extreme Learning Machine(ELM)and Partial Least Squares Regression(PLS),were used to detect the content of soluble solids in apples.After the ELM algorithm creates the network,the training set samples are applied to verify the network classification results,and then the test set samples are predicted.The test results of PLS and ELM are in line with the corresponding expectations.In order to achieve precise positioning and overcome the traditional apple classification model’s requirement for the accuracy of the original value,it is necessary to solve the problem of the accuracy of the training sample class target.This paper also introduces uncertainty into the classification problem,that is,based on the DS evidence theory apple classification level fusion algorithm,the ELM soluble solids content prediction model and the PLS soluble solids content prediction model are combined with DS evidence theory.The classification accuracy rate reaches 94.697%,and the classification accuracy after using DS fusion is greatly improved compared with the single method classification so as to solve the problem of the decrease of the classification accuracy rate caused by hard segmentation.