Study On Quality Detection Method Of Korean Pine Seed Using Machine Vision And Near Infrared Spectrum

Pine nuts are very valuable in forestry economy.China is a large producing country of pine nuts,and Korean pine seeds in the Northeast are the most famous.The market of pine nut in our country is very wide,which is 60 percent to 70 percent of the volume of trade in the whole world.Pine nuts,as one of the economic crops,can bring good economical benefits.Recently in our country,the output value of pine nuts be processed is increasing,but it is still lower then the nature value of pine nuts.After commercial processing,pine nuts can make more benefit,which provides wide application prospects for the nondestructive testing of the quality of pine nut.The demands of the quality supervision of pine nut and the development and utilization of pine nut and the further processing of pine nut in the market of pine nut in our country promotes the development of the nondestructive testing of the quality of pine nuts.Recently the study on nondestructive detection method of pine nut hasn’t been carried out widely.Especially the study on internal quality inspection of fat and protein and moisture in pine nut hasn’t been carried out yet.Manual selection and grading of mechanical vibration are usually used to grade the external quality of pine nut.Manual selection needs a lot of labor force,and the labor intensity is very heavy,and the results of grading are affected by the subjective factors.The noise and the power consumption of vibration grading is heavy and high,and the accuracy of grading is not so good because of the large difference between grades.Meanwhile grading of mechanical vibration can make pine nuts bump together.The method of physicochemical analysis is usually used to inspect the internal quality of fat and protein and moisture in pine nut.The method of physicochemical analysis is complex,expensive and off-line,and it consumes much time.How to establish an external quality graduation and internal quality inspection method which is fast,accurate,easy,safe and nondestructive is a worthy and urgent question.In such background,in this dissertation,Korean pine seed with peel and without peel is set to be the object,to carry out the research of nondestructive graduation of external quality of Korean pine seed based on machine vision and the research of nondestructive inspection of internal quality of Korean pine seed based on near infrared spectroscopy.On the side of the nondestructive graduation of external quality,Korean pine seed with peel is set to be object.To promote accuracy and detection speed of the extraction of Korean pine seeds’ contour,based on the traditional C-V model,the improved C-V model is studied to extract the contour of Korean pine seed.On this basis,Korean pine seeds’ length and the maximum width are extracted.The mathematical models of Korean pine seeds’ length and the maximum width are established.It finds that the mean prediction accuracy of the length model is 98.42%,and that of the maximum width model is 96.10%.The comprehensive evaluation standard for external quality graduation of Korean pine seed is proposed based on Korean pine seeds’ length and the maximum width.and the mean accuracy of graduation is 97.2%,which shows the reliability and accuracy of graduation.On the side of the nondestructive inspection of internal quality of Korean pine seed,Korean pine seed with peel and without peel are set to be the objects.Because the modeling accuracy of derivative preprocessing is impacted by window widths.Therefore confirmation of window widths of first derivative(1-Der)and that of second derivative(2-Der)of fat,protein and moisture in Korean pine seed is studied respectively.It finds that window width of 1-Der and that of 2-Der are both 5 that the accuracy of models of fat,protein and moisture in Korean pine seed with peel is more desirable.Window width of 1-Der and that of 2-Der are 10 and 25 respectively that the accuracy of models of fat,protein and moisture in peeled Korean pine seed is more desirable.Because the modeling accuracy of backward interval partial least squares(BiPLS)is impacted by partition numbers.Therefore selection of partition numbers of BiPLS of fat,protein and moisture in Korean pine seed is studied respectively.It finds that the accuracy of models of fat,protein and moisture in Korean pine seed with peel is better when partition number is 15.The accuracy of models of fat,protein and moisture in peeled Korean pine seed is better when partition number is 10.To establish high-quality and nondestructive near infrared(NIR)models of fat and protein and moisture in Korean pine seed.The influence of different spectral preprocessing methods(1-Der,2-Der,Multiplication Scatter Correction(MSC),Vector Normalization(VN),Standard Normalized Variate(SNV))and that of the different bands selection methods(Interval Partial Least Squares(iPLS),Backward Interval Partial Least Squares(BiPLS),Elimination of Uninformative Variables(UVE))on the modeling accuracy is studied,to determine the comparatively better spectral preprocessing method and bands selection method and the appropriate modeling wavelength range.It shows that the fat and protein and moisture models of Korean pine seed with peel are comparatively better after VN and BiPLS.The protein model of peeled Korean pine seed is comparatively better after SNV and BiPLS.The fat and moisture models of peeled Korean pine seed are comparatively better after 1-Der and BiPLS.The root-mean-square error of validation(RMSEP)of the optimal model of fat and protein and moisture of Korean pine seed with peel is 0.7651,0.6670,1.0417 respectively.The RMSEP of the optimal model of fat and protein and moisture of peeled Korean pine seed is 0.6468,0.5761,0.8338 respectively.The quantitative nondestructive inspection of internal quality of fat and protein and moisture in Korean pine seeds is realized.A new method and approach of nondestructive graduation of external quality and nondestructive inspection of internal quality of Korean pine seed are improved by the realization of this study.It also has certain application value to the research of quality evaluation of other nuts.