Research On Near Infrared Spectroscopy Detection Method Of Hami Brix And Design Of Portable Device

Hami melon is a collective name for the thick-skinned melon varieties in Xinjiang,and the flesh is rich in nutrients.Sugar content is one of the important indicators for evaluating the quality of Hami melon.If it is overripe and not picked,it will cause the loss of moisture and nutrients inside the Hami melon,resulting in cracks or even hollowness.The unripe Hami melon will have a dry taste and low sugar content.The visual inspection method has low efficiency in judging the quality of Hami melon,and the recognition accuracy fluctuates with the subjectivity of the person.In addition to the visual inspection method,destructive testing is the main method.Currently,most of the online fruit sugar detection systems developed are applied to postharvest.In response to the above problems,this research is based on the nearinfrared spectroscopy non-destructive testing technology,selecting appropriate chemometric methods and external parameters,exploring multi-feature fusion to construct a prediction model of Hami melon sugar content,and designing an Android-based mobile phone based on the above theoretical research Portable non-destructive testing device for Hami melon sugar content.The main contents of this research are as follows:1.The relationship between texture characteristics of Hami melon and sugar content.Statistical analysis of the physical parameters and sugar content of the sample yielded P mass,Showed that the collected physical parameters have no significant linear correlation with the sugar content,so it was not feasible to predict the sugar content based on the physical parameters of the sample surface.Based on the GLCM,the texture features of Ent,Asm,Con and Corr are extracted from 4 angles(0°,45°,90°,135°).In order to improve the detection speed,the selected texture features and sugar content are extracted.Correlation analysis showed that Ent,Asm,Corr and sugar content were all positively correlated,and the correlation coefficients are all above 0.5,so the degree of correlation was relatively large;while contrast was negatively correlated with sugar content,indicating that contrast is a texture feature and the degree of correlation between sugar content Smaller.Therefore,Ent,Asm,and Corr are selected as the texture features of the Hami melon samples to provide a certain theoretical basis for the subsequent multi-feature fusion modeling.2.Multi-feature fusion constructs a prediction model of Hami melon sugar content.Comparing a variety of spectral preprocessing and characteristic variable screening methods to construct a prediction model of Hami melon sugar content,the results showed that SG combined with MSC and FD has the best effect on Hami melon spectral preprocessing(rp=0.8155,RMSEP=0.8367);SPA characteristic wavelength screening is linear and Nonlinear modeling is the best method;by comparing the accuracy of the prediction model of PLS and LSSVM,the prediction accuracy and robustness of the Hami melon sugar content model established by PLS is the best(rp=0.8861,RMSEP=0.7734),of which the SPA-PLS model Compared with the full-spectrum model,the rp was increased by 5.31%,and only 2.36% of the wavelength information of the entire spectrum is used;for the identification of singular samples,the Mahalanobis distance method combined with Monte Carlo cross-checking can better identify the singular samples in the data.Accurate(rp=0.9186,RMSEP=6911);for multifeature fusion modeling,spectral information and texture information fusion modeling effect is the best(rp=0.9227,RMSEP=0.5750),better than using only spectral feature modeling(r P=0.9186,RMSEP=0.6911),which was better than only using texture feature modeling(rp=0.6706,RMSEP=0.9124).Provide a high-precision sugar content prediction model for the subsequent design of portable devices.3.The influence of different external parameters on the near-infrared spectroscopy model and the hardware design of portable devices.Collect the corresponding spectral information under different light source angles and object distance height conditions,and establish PLS and LSSVM sugar prediction models.The modeling results under different parameter conditions showed that the modeling results are the best under the conditions of the light source incident angle of 40° and the object distance height of 170mm(rp=0.6227,RMSEP=1.2850).Design the hardware part of the portable testing device for sugar content of Hami melon based on Android system.4.The software design of a portable non-destructive testing device for Hami melon sugar content.Based on Android studio platform development,the software part includes login and registration interface,positioning module,data processing module,etc.