Internal Quality Nondestructive Testing Methods Of Fruits Based On Dielectric Property And Spectrum Technology

Fruits’development process is divided into three stages:growth,mature and aging.Growth stage and mature stage were also known as growth or growing period.During this period,fruits’internal structure will be changed a lot.At present,more research on characteristics of fruits has focused on postharvest and storage period.Researchers more rarely focused attention on fruits’quality change rule and characteristics in the growing period.Changes and monitoring of the quality of fruits in the growing process will directly affect the quality of fruits,fruit varieties breeding,and ultimately affect fruits’sales,storage and subsequent processing.Therefore,it is necessary to do further research on the characteristics of fruits in growing period.In addition,many tests for fruit’s properties were based on multiple indexs studies by using a single detection method.There are few comparative study based on several methods for a kind of fruit.Aiming at these two problems,the following research was carried out:（1）Research on the relationship between the internal quality of fruits and the rules of quality change in the process of fruit growthThe dielectric spectrum,the near infrared spectrum and high spectrum methods were selected as the testing means.The growth period"Fuji"apples and growing"Dangshansu"pears were measured as the objects.The relationship between the physical and chemical properties of these fruits and parameters of their internal quality were analyzed.The results showed that with increased development time,dielectric parameters of"Fuji"apples and"Dangshansu"pears overall all showed a trend of decrease.Combined with the changes in the release and respiration of the fruits ethylene,samples of"Fuji"were judged to meet the harvest condition at the 167^th day from their development time,while"Dangshansu"pears were judged at the143^rd day.Therefore,by detecting dielectric properties of"Fuji"apples and"Dangshansu"pears,combined with growth parameters such as the hardness and sugar in the process of change,the maturity of the fruits could be predicted.（2）Research on the likelihood for predicting parameters of fruit’s internal quality based on the dielectric parameters at a single frequency or the near infrared absorbance intensity（or hyperspectral reflectance）at a single wavelength,based on the dielectric spectrum,the near infrared spectrum and the high spectrum technologyThe dielectric properties（20⁴500MHz）,near infrared spectral characteristics（833²500nm）and high spectral characteristics（900¹700 nm）of the growth period"Fuji"apples,"Dangshansu"pears,ripe melons and fragrant pears were measured.The linear correlation between the dielectric parameters（at a single frequency）,or the near infrared absorbance intensity（at a single wavelength）,or the hyperspectral reflectance（at a single wavelength）,and the fruit internal quality parameters were analyzes.The results showed that there were the relatively stable linear correlation between dielectric parameters and the sugar,the hardness of"Fuji"apples in the range of 100¹000MHz frequency.The linear relationship between moisture content,pH value and the dielectric parameters were weaker.The linear correlation between the dielectric parameters,the near infrared absorption intensity（or the high spectral reflectance）of other experimental fruits and their internal quality parameters were weak,too.Therefore,the dielectric parameters at several certain frequencies in the range of100¹000MHz could be used to predict the sugar or the hardness of the"Fuji"apple.It’s difficult to predict the other quality parameters of"Fuji"apples and all the internal quality of"Dangshansu"pears by the characteristic parameters at a single frequency.It’s also difficult to predict the internal quality of the growing"Fuji"apples,"Dangshansu"pears,ripe melons and fragrant pears by using the light intensity（or high spectral reflectance）at a single wavelength.（3）Models design for fruit internal quality parameters testing based on the dielectric spectrum technologyGrowth period"Fuji"apples and ripe melons selected as the research objects,testing models for fruit internal quality were designed.Moreover,the influence on the model prediction effect of different modeling methods and variables optimization methods were analyzed.The results showed that the quality parameters of"Fuji"apples and ripe melons could be predicted reasonably based on dielectric spectrum technology.The prediction effect of apple’s sugar content and moisture content was better than that of melon.（4）Models design for fruit internal quality parameters testing based on the near infrared spectroscopyGrowth period"Fuji"apples,growing"Dangshansu"pears,ripe melons and fragrant pears selected as the research objects,testing models for fruits internal quality were designed.Moreover,the influence on the model prediction effect of different modeling methods and variables optimization methods were analyzed.The results showed that the quality parameters of all fruits could be successfully predicted based on the near infrared spectroscopy technology.The prediction effect of fragrant pears was best and melons was worst.（5）Models design for fruit internal quality parameters testing based on the hyperspectral technologyGrowth period"Fuji"apples and growing"Dangshansu"pears selected as the research objects,the internal quality testing models were designed and studied.The results showed that the detection effect of apples was better than that of the pears on sugar content detection.In the aspect of hardness testing,the detection effect of pears was better than that of apples.For the moisture content detection,the detection effect of the pears was better than that of apples.（6）Comparison of different detection methods and models for each quality parameter of fruitsIn order to seek the best detection method and the optimal prediction model for different internal quality parameters of all fruits from all testing models based on the dielectric spectrum,the near infrared spectrum and the high spectrum methods,the prediction effect of the same testing parameter belonged to different fruits were compared and analyzed.The results showed:1)The best detection method for sugar and moisture content of growing"Fuji"apples both were near infrared spectroscopy,and the optimal detection model was SPA-ELM and PCA-ELM,RMSEP was 0.444°Brix and 0.624%,repectively.Hardness and pH value were dielectric spectrum method,and the optimum detection model was UVE-LSSVM and SPA-ELM,RMSEP was 0.447 kg·cm^-22 and 0.005,respectively.2)The best detection method for sugar and hardness of"Dangshansu"pears in the growing period both were near-infrared spectroscopy,and the optimal detection model both were UVE-LSSVM,RMSEP was 0.271°Brix and 0.399 kg·cm^-2,respectively.The best detection method of moisture content was hyperspectral method,and the optimal detection model was PCA-ELM model,RMSEP was0.564%.3)The best detection methods for sugar and moisture content of ripe melons both were dielectric spectrum method,and the optimal detection model was SPA-LSSVM and UVE-LSSVM,RMSEP was 1.096°Brix and 1.404%,respectively.4)The inner quality parameters of fragrant pears were detected by near infrared spectroscopy.The optimal model of sugar detection was UVE-LSSVM,RMSEP was 0.252°Brix;and the optimum model of hardness and moisture content both were PCA-LSSVM,RMSEP was 0.348 kg·cm^-2and 0.329%,respectively.