Research On Nondestructive Testing Of Hardness,Moisture And Soluble Solids Content Of Apple Based On Hyperspectral Imaging Technology

Apple contains a variety of minerals and vitamins.It is one of the most popular fruits.China is the world’s largest producer and consumer of apple.With the rise of people’s living standards,the quality requirements for apple has gradually increased.Apple’s quality includes two aspects:internal and external.External qualities include color,size and appearance.Internal qualities include sweetness,moisture and hardness.Currently in our country’s fruit market,apple’s quality varies greatly.Which is caused by the backward picking grading detection technique.In recent years,nondestructive detecting of fruit’s internal quality based on hyperspectral imaging technology has become a hot research topic.This technique combines the image information and spectral information of the target,which can be used to analyze the internal and external quality of target rapidly and nondestructively.In this paper,hyperspectral imaging technology was used to nondestructive detect apple’s hardness,moisture and soluble solids content.The main research contents are as follows.(1)Hyperspectral imaging system from 400nm to 11 OOnm was used to detect the content of soluble solid content in apple.Black and white correction for the spectral images collected by the system.A circular region of interest of 150 pixels was selected.The collected spectral data were pretreated by S-G first order differential.The SPA algorithm was used to extract 12characteristic wavelengths from the pretreated data.(740.86、752.95、785.99、800.34、813.59、835.70、842.34、860.05、883.30、897.71、938.70 and 950.89nm).The BP and GA-SVR prediction models were established respectively.The experimental results showed that the GA-SVR model was better than the BP neural network model.The correlation coefficient and root mean square error of the calibration set and the prediction set were 0.8806,0.2607 and 0.8505,0.3031 respectively.(2)Hyperspectral imaging technology was used to detect the moisture content of apple.PC A and SPA algorithm were used to extract the characteristic wavelength of the pretreated spectral data.The cumulative contribution rate of the first 7 principal components of PCA algorithm had reached more than 95%.SPA algorithm extracted 16 characteristic wavelengths.PSO and Grid Search algorithm were used to optimize the SVR parameters.PCA-PSO-SVR,PCA-Grid-SVR,SPA-PSO-SVR and SPA-Grid-SVR four moisture prediction models were established.The experimental results showed that the SPA-Grid-S VR model was the best one.The correlation coefficient and root mean square error of the calibration set and the prediction set were 0.9132,0.2236 and 0.8754,0.2387 respectively.(3)Hyperspectral imaging technology was used to detect the apple flesh hardness.The spectral data after pretreatment,14 characteristic wavelengths were extracted by SPA algorithm.GA algorithm was used to optimize the SVR and BP parameters.The GA-BP prediction model was established by using the optimized BP network weights and thresholds.At the same time,the GA-SVR prediction model was esta’blished by using the optimized parameters c and g.The experimental results showed that the prediction error of GA-BP model was less than that of BP network.GA-SVR model was better than GA-BP model.The correlation coefficient and root mean square error of the calibration set were 0.8543 and 0.3203.The correlation coefficient and root mean square error of prediction set were 0.8135 and 0.2967.