| Grapes are hailed as the first of the four major fruits.In 2017,China’s total output reached 13.08 million tons.Red globe grape is one of the grape varieties.It is widely loved by people because of its full granules,firm flesh,sweet and delicious,and rich in nutrients.With the improvement of people’s living standards,while paying attention to the appearance quality of red globe grape,consumers prefer the internal quality with good taste and good flavor.At present,the traditional destructive testing methods are mostly used for the internal quality testing of red globe grape.Because the testing methods are tedious and time-consuming,and only sampling testing can be carried out,the testing range is small,and the tested samples are completely damaged and cannot be sold,resulting in greater waste.In this paper,red globe grape extract from different growing periods is taken as the research object,the internal quality of red globe grape extract from growing periods is studied by using visible/near infrared spectroscopy,hyperspectral imaging and other technologies,and a portable internal quality detector for red globe grape extract fruit grains and red globe grape extract strings is designed.Based on various spectral image technologies,a non-destructive detection model for the content of soluble solids in red globe grape extract by hyperspectral image information fusion is established to further improve the prediction accuracy of the detection model.The main research contents and conclusions are as follows:1)Study on the variation law of internal quality parameters and spectral characteristics of red globe grape grains and red globe grape strings in growing period.Using visible/near infrared spectroscopy and hyperspectral imaging detection technology,the change law of internal substance content(soluble solids,total acid,p H,firmness and moisture content)of red globe grape extract during the growing period with the gradual ripening of fruits was studied.The corresponding relationship and change process between visible/near infrared spectral characteristics and hyperspectral characteristics at the growth and maturity stage of red globe grape extract and internal quality were analyzed,and the correlation between internal substances in the growth stage was analyzed.The results showed that the changes of souble solids content(SSC)and p H value of red globe grape extract were the same during the whole growth period of red globe grape extract,and the index value gradually increased with the maturity of red globe grape extract.The change trends of total acid,firmness and moisture content of red globe grape extract are the same,and all indexes gradually decrease with the maturity of red globe grape extract.During the growth process,the total acid and moisture content have the same change trend.From the 61 st to 81 st days of growth and development,both of them drop sharply.From the 91 st to 116 th days of growth and development,the total acid and moisture content basically remain at a fixed value with little change fluctuation.Souble solids content are highly correlated with p H,total acid,firmness and moisture content.Under the near infrared diffuse reflectance spectrum,the positions of peaks and troughs of red globe grape fruits are relatively fixed.In the early stage of growth,the distance between the two curves is relatively large in the same growth period.As the fruits mature gradually,the distance between the two curves gradually decreases in the same time interval.In the visible/near infrared spectral band,the spectral curve of the spectral band in 550 nm to 600 nm changes greatly during the growth process of red globe grape,and the position of the peak in this interval will gradually change from 545 nm to 635 nm.In the early stage of growth,the distance between the two curves is relatively large in the same growth period.As the fruit matures,the distance between the two curves gradually decreases in the same growth period.2)Research on the internal quality detection model of red globe grape fruit kernels based on visible / near infrared spectroscopy.Taking the red globe grape-extracted fruit grains in the growing period as the research object,the visible/near infrared diffuse reflectance spectrum acquisition system of the red globe grape-extracted fruit grains is established by using Antaris II Fourier transform near infrared spectrometer and USB2000 spectrometer respectively,and the visible/near infrared transmission spectrum acquisition system of the red globe grape-extracted fruit grains is established by using Maya2000 pro spectrometer.The internal quality prediction models of the red globe grape-extracted fruit grains in the transmission and diffuse reflectance modes are established respectively,and the advantages and disadvantages and applicable scenes of the models under the transmission mode and diffuse reflectance mode are compared and analyzed.The influence of placement mode on the transmission spectrum acquisition device built by Maya2000 pro was studied.The results show that the contents of soluble solids,total acid,p H,firmness and moisture content of red globe grape extract can be accurately predicted under diffuse reflection mode and transmission mode.Under the same conditions,the model built by collecting spectra through transmission mode is better than that built by collecting spectra through diffuse reflection mode.Under the Maya2000 pro optical fiber spectrum acquisition system,the placement mode has great influence on the modeling.For SSC,total acid,p H value,firmness and moisture content indexes of red globe grape extracted fruit grains,the prediction effect of the model built by average spectrum is the best,and the prediction effect of the model built by vertical mode is better than that of horizontal mode.3)Research on internal quality detection model of red globe grape string based on visible/near infrared spectroscopy.In order to solve the problems of complex appearance and different sizes of red globe grape string,it is difficult to collect the spectrum of red globe grape string.Taking the red globe grape string in growing period as the research object,the 3D modeling of the spectrum collection probe is carried out by UG,and the spectrum probe collection device is manufactured by 3D printing technology.Combined with the spectral probe acquisition device,the diffuse reflection spectral acquisition method is adopted to build the spectral acquisition device of the red globe grape string.A diffuse reflection internal quality prediction model of red globe grape string was established.The prediction accuracy of the model was improved through a series of methods such as spectral pretreatment and feature wavelength extraction.An optimal model and detection method for internal quality prediction of red globe grape string were determined.The results show that the RPD of the optimal PLSR model for total acid and p H of the red globe grape extraction string total acid and p H set up by the set-up device are all greater than 2.5,indicating high prediction accuracy,and the RPD of SSC is 2.2501 model with certain predictability.The PLSR optimal model established by red globe grape stripe string has certain differences in the prediction performance of each index.The prediction performance of p H is the best,followed by total acid,and SSC is relatively the worst,but the prediction performance can meet the accuracy requirements.The effect of the optimal model of SSC,total acid and p H established by red globe grape stripe string is obviously lower than that of the model established for red globe grape stripe fruit grains.The red globe grape extract string detection model reflects the whole information of the whole red globe grape extract string,and can effectively measure the internal substances of the whole red globe grape extract string.The established optimal red globe grape model can be applied to monitor the internal quality index during the growth process of red globe grape in the field and has important practical significance.4)Research on internal quality detection model of red globe grape extract fruit grains based on hyperspectral imaging technology.In order to study the prediction effect of hyperspectral imaging technology on the internal quality of red globe grape-extracted fruit grains and analyze the differences between the models established by hyperspectral imaging technology and visible/near infrared technology,taking red globe grape-extracted fruit grains in growing period as the research object,based on hyperspectral imaging technology,the influence of different characteristic band extraction methods on the prediction effect of the model is compared through different spectral preprocessing methods,different prediction models are established,the advantages and disadvantages of the different prediction models are compared and analyzed,and an optimal prediction model is determined.For SSC index,based on spectral information,image information and fusion of spectral information and image information,three different feature information prediction models are established respectively.The results showed that for SSC,total acid,p H,firmness and moisture content indexes of red globe grape extract,the effect of model with fruit stalk horizontally placed was better than that with fruit stalk side up and fruit stalk side down,and the effect of model with fruit stalk side down was better than that of model with fruit stalk side up.The model based on the average spectrum of the whole fruit grains has the best effect.The best model of total acid of red globe grape extracted fruit grains is the method of secondary wave band extraction.The best models of SSC,p H,firmness and moisture content of red globe grape extracted fruit grains are all the methods of primary wave band extraction.Through hyperspectral imaging technology,internal indexes of SSC,total acid,p H,firmness and moisture content of red globe grape extracted fruit grains can be accurately predicted.The prediction performance of SSC,total acid and p H model is obviously better than that of firmness and moisture content model.For PLSR and LSSVM models,under the same characteristic wavelength extraction,the effect of nonlinear LSSVM model is better than that of linear PLSR model,and the model is more stable,but the operation time of the model is longer,and the accuracy of PLSR prediction model is better,which can meet the detection of internal quality of red globe grape.The correlation coefficient of the model based on spectral features is larger than that of the model based on image features.Compared with image information,spectral information can better predict SSC.The model built after image spectral fusion is the best,and the model has the best stability.The effect of LSSVM model built is better than that of PLSR model.5)Selection of optimal detection model for single quality index of red globe grape and determination of comprehensive quality index.In order to deeply analyze the prediction advantages and disadvantages of different detection methods for each index and obtain the optimal detection method and the optimal prediction model,aiming at each internal quality detection index of red globe grapeextracted fruit grains and red globe grape-extracted strings,the application places and characteristics of the optimal models established under various methods are analyzed and compared,and the indexes of the measured red globe grape-extracted fruit grains and red globe grape-extracted strings are fused through factor analysis method,so that the comprehensive quality indexes of the red globe grape-extracted fruit grains and red globe grape-extracted strings in the growing period under the visible/near infrared diffuse reflectance spectrum acquisition mode are established,and the overall evaluation of the internal quality of the red globe grape-extracted fruit grains and red globe grape-extracted strings is realized.The results showed that the optimal prediction model of SSC and moisture content of red globe grape-extracted fruit grains in growth period established by near-infrared diffuse reflectance spectroscopy had the best effect,and there was more information related to SSC and moisture content index of red-extracted fruit grains in this band range,so the prediction performance was better.The optimal prediction model of total acid,firmness and p H of red globe grape extracted fruit grains in growth period established by visible/near infrared transmission spectrum is better than that established by visible/near infrared diffuse reflection spectrum.The total acid,firmness and p H indexes of red globe grape extracted fruit grains in growth period can be better predicted by transmission spectrum acquisition.It is feasible to use spectral technology to detect red globe grape fruits and red globe grape strings in growth period.6)Research and development of portable nondestructive testing device for internal quality of red globe grape-extracted fruits and red globe grape-extracted strings.Based on the size and characteristics of the red globe grape-extracted fruit grains and red globe grape-extracted strings in the growing period,the spectral probe acquisition devices of the red globe grape-extracted fruit grains and red globe grape-extracted strings are designed respectively,and a portable detector for the red globe grape-extracted fruit grains and red globe grape-extracted strings is built.The above-mentioned models are implanted into the instrument to realize the comprehensive detection and classification of the mature state and internal quality of the red globe grape-extracted fruit grains and red globe grape-extracted strings.The results showed that the SSC,total acid,p H,firmness and moisture content of the red globe grape-extracted fruit grains could be detected simultaneously by the portable red globe grape-extracted fruit grain internal quality detector,and the SSC,total acid and p H internal indexes of the red globe grape-extracted fruit grains could be detected simultaneously by the portable red globe grape-extracted fruit grain internal quality detector. |
PDF Full Text Request