Study On The Detection Technology Of Fruit Quality Based On Dieleciric Spectroscopy

In order to investigate the feasibility of dielectric nondestructive detection of fruit quality based on the electrical spectrum.The dielectric parameters of strawberry,Long jujube under of 55 frequency points in the 1kHz?1 MHz frequency range using HIOKO-3532-50 LCR parallel plate measuring instrument were measured and analyzed combining with quality index analysis after the results are as follows:Study respiration,soluble solids and weight loss rate varied regulation of 50 strawberries during storage period of 5 days;strawberry dielectric parameters and quality indexes were analyzed by correlation and regression stepwise regression method;GA method was used to filter out the characteristic frequencies for respiration strength,soluble solids content and weight loss rate of strawberry;the PLS quality prediction model was established by dielectric parameters of characteristic frequency;Dynamic changes were analysed under of 630.96kHz(105 85Hz),707.95kHz(105.90Hz),891.25kHz(105 95Hz)and 1000kHz(106Hz);the two fitting relationship between quality indexes and ?'and the strawberry quality optimization model using SVM method was establish.Results show:(1)Strawberry dielectric parameters and quality indexes were analyzed by correlation and regression stepwise regression method.The results show that the 8 preferred frequency points(1.00kHz?15.85 kHz?35.48 kHz?39.81 kHz?501.19 kHz 1000kHz the relative dielectric constant of'and 446.68 kHz?707.95kHz the dielectric loss factor ?")has a strong correlation with the dielectric parameters of strawberry quality index,reached significant level(p<0.01).(2)34?39?5 characteristic frequency points were screened by genetic algorithm(GA),Strawberry PLS quality prediction model was established by the dielectric parameters of the characteristic frequency above.The results show that:R2 value of GA-PLS model of respiratory intensity,soluble,weight loss were respectively 0.94,0.85,0.91,and measured values had no significant difference in prediction of the quality indexes for each quality index(p>0.05).(3)Dynamic changes of ?' were analysed under of the 630.96kHz,707.95kHz,891.25kHz and 1000kHz and quality index optimization models were established using SVM method.The results show that it has a strong correlation between the 4 preferred test frequencies of ?' and strawberry quality indices.SVM optimization model decision coefficient R2 of Strawberry respiration,soluble solids,the weight loss rate were 0.94,0.92 and 0.81;Compared with GA-PLS,SVM optimization model of the required frequency points less,while the model effect is consistent.100 jujubes during the storage period of 10 days as the research object,its characteristic frequencies were screened out by competitive adaptive reweighted sampling(CARS),genetic algorithm(GA)and successive projections algorithm(SPA),uninformative variable elimination(UVE).Besides,least squares(PLS)and least squares support vector machine(LSSVM)prediction model were established by combining Vc value with its characteristic frequencies above.The results show that 23,35,14,38 characteristic frequency points were slected by CARS method,GA method,SPA method,UVE method from full spectrum variables for long jujube full spectrum;The PLS the LSSVM prediction model of Changzao Vc were established by combining characteristic frequencies with Vc values,the model effect was analyzed that UVE algorithm is the best spectrum screening method;UVE-PLS is the best prediction model,the Rc,RMSEC,Rp,RMSEP values were 0.9871,3.9322,0.9460,4.0400,verifying model R2 value is 0.9617.