| In order to obtain Apple’s internal quality information conveniently and effectively,to achieve apple mold heart disease detection and quality classification,and to provide theoretical support for Apple’s nondestructive testing.The study of non-destructive testing of apple mold heart disease detection and apple internal quality grading methods has important practical significance and broad application prospects for improving the level of apple quality testing and grading,ensuring the rapid development of the apple industry,and highlighting the advantages of high-quality apples.Aiming at the problem of apple mold heart disease detection and quality grading,this paper takes apple dielectric and physical and chemical indicators as the research objects,uses standardization for data preprocessing,and uses different modeling methods to construct an apple mold heart disease non-destructive test model and an apple quality grading model.The main work of this paper is as follows:(1)In view of the data requirements for constructing an apple mold heart disease non-destructive test model and an apple internal quality grading model,Collected 12dielectric parameters of Apple at 9 frequencies and six physical and chemical indicators such as soluble solids,titratable acid and hardness as the research objects,In view of the uneven distribution of the data due to the different frequency and dimension of the dielectric characteristics,the Z-score method is used for preprocessing to eliminate the unevenness of the dimension and data distribution.Meanwhile,in order to eliminate the noise and redundant information in the original data,this paper proposed an improved principal component analysis algorithm based on the re-weighting method,so that it can have more sparse solution,thereby improving the experimental effect.(2)Aiming at the problem that apple mold heart disease cannot be effectively identified by appearance,in order to achieve quick and effective detection of apple mold heart disease,this paper uses 108 dielectric parameters of 220 apples as the research object,and uses Z-score method,principal component analysis algorithm and improved principal component analysis algorithm for data preprocessing.The construction of mold heart disease fruit detection model based on BP neural network,support vector machine and random forest algorithm is proposed respectively,and the model accuracy of two dimensionality reduction algorithms and three modeling methods are analyzed.The experimental results show that the improved dimensionality reduction algorithm and the model constructed by random forest have an accuracy of 98.00%and 97.14%in the training set and test set,respectively.(3)the traditional apple internal quality testing methods complicated,expensive equipment and other issues,this article is the realization Apple Internal Quality grading convenient economy,proposed the apple quality classification method based on dielectric characteristics.For questions apple grading standards difficult to define,based on the existing data will Apple be divided into three levels.This paper uses 108 dielectric and 6physical and chemical data of 110 apples as the research object.Based on correlation analysis and multiple regression model,the dielectric index-physical and chemical index regression model is constructed,and the soluble solids,titratable acid,hardness,and media are constructed.For the regression model of the electrical indicators,theR~2 values are0.951,0.94,and 0.922.Finally,a support vector machine was used to build an Apple quality grading model,and the classification accuracy rates in the training set and test set reached93.5%and 90.1%. |
