The Study Of On-line Nondestructive Detecting And Sorting System Of The Quality Of Fresh Jujube Based On Near Infrared Spectroscopy

Jujube originated in china, and its total output has occupied the first place in the world, but the volume of exports is a few, the main reason is because of fresh jujube dates and online testing technology level is relatively backward, low level of commercialization of fresh jujube dates. Fresh jujube quality testing sorting is to promote value-added fresh jujube dates, fresh jujube dates to improve one of the main methods of industry profits rise, sorting jujube can promote the commercialization and brand development by Fresh jujube quality inspection, therefore, to achieve within the external fresh jujube dates quality testing sorting is very important and necessary. Currently, the online home and abroad for the quality of non-destructive testing fresh jujube sorting also reported less relevant NDT techniques and sorting method is not mature enough, there are no external quality within fresh jujube line non-destructive testing equipment, and a few foreign fruit quality sorting line detection equipment because of its expensive purchase price and maintenance, can not be a wide range of popular applications.In this study, jujube as the research object, soluble solids and minor injuries within fresh jujube Quality of external indicators, designed jujube quality online spectrum NDT sorter; Using near infrared spectroscopy technology, the establishment of jujube internal quality prediction models and external damage discriminant model combined with chemometrics methods, and the model to analyze the accuracy and stability compared to the preferred method of modeling the best of the last verify the feasibility and accuracy of the program through an online detection device. This paper made a study of the following aspects:(1) Fresh jujube on the internal quality indicators-soluble solids extracted wavelength and prediction model based on visible/near-infrared feature. Spectral band in the whole range of spectral data through different pretreatment methods after the establishment of PLS quantitative analysis model to determine the best method of pretreatment and 9 o'clock smooth; then, after 9:00 smoothed data preprocessed RC, SPA, characteristic wavelength CARS three methods of extraction; Finally, each extracted feature data and a full wavelength band data were established PLS model and LS-SVM model, and analyze the accuracy of the model is established, and ultimately select the RC-LS-SVM method is most best Winter SSC sample value prediction model, where RC is 0.9347, RMSEC to 0.9928, RP is 0.9176, RMSEP to 1.0828.(2) Fresh jujube on the internal quality indicators-soluble solids extracted wavelength and prediction model based on visible/near-infrared feature. Spectral band in the whole range of spectral data through different pretreatment methods after the establishment of PLS quantitative analysis model to determine optimal pretreatment method-MSC; then, the data through MSC preprocessed RC, SPA, CARS three kinds the method of extracting characteristic wavelength; Finally, each extracted feature data and a full wavelength band data were established PLS model and LS-SVM model, and analyze the accuracy of the model is established, and ultimately select the RC-LS-SVM method is the best jujube sample SSC value prediction model, where RC is 0.9441, RMSEC to 1.0175, RP is 0.9363, RMSEP to 1.1916.(3) Fresh jujube appearance of quality indicators-slight damage in the visible wavelength extraction and forecast near-infrared feature/model. Spectral band in the whole range of spectral data through different pretreatment methods after the establishment of PLS quantitative analysis model to determine the best method of pretreatment and 9 o'clock smooth; then, after 9:00 smoothed data preprocessed RC, SPA, characteristic wavelength CARS three methods of extraction; Finally, each extracted feature data and a full wavelength band data were established PLS model and LS-SVM model, and analyze the accuracy of the model is established, and ultimately select the RC-LS-SVM method is most good jujube sample minor injuries discriminant, discriminant accuracy of 97.50%.(4) Fresh jujube quality online Nondestructive detection and sorting accuracy verification. With the present system Design Institute of fresh jujube samples of different soluble solids and minor damage fresh jujube samples were detected sorting, in order to verify the accuracy of detection sorting apparatus. The results show that the system is designed in this study plan is feasible and can achieve high sorting accuracy.