| China is a country with vast amount of fruit production.There are rich in resources of fruit tree and long history of tree cultivation.Since 1993,output of fruit of our country and cultivated area have been occupying first place in the world steadily all the time.In recent years,though the output of fruit of our country is increasing constantly,fruit export quantum is still very small,its reason is that commercialized degree of fruit is low,the slightly bad and good green bristle-grass of quality mixes,cause the competitiveness in the international market to be weak.Backwardness of treatment technology after it is adopted that an important reason for influencing the commercialized degree of fruit and competitiveness of our country is the fruit,the hierarchical ability of fruit especially of our country is weak,it is low to measure efficiency.Traditional fruit quality differentiation and hierarchical method are time-consuming and strenuous,apt to receive the interfering of human factor(such as the sense of smell,sense of taste,taste),and lack the objective and rational discrimination standard to the nutritional labeling within the fruit.And measure one by one after the chemical analytic approach needs to carry on the fruit brokenly,the sample that is often measured is limited in quantity,the representative difficult to guarantee to sample,thus unable to realize that measure harmlessly fast.The assembly line is measured and grade the equipment to usually only grade roughly according to the size of the fruit or the weight after some current fruit is picked,and the expecting relatively much special-purposely of equipment,the utilization ratio is low.So,develop a kind of fast,high-efficient fruit detection technique can't harming,in order to meet the need of fruit extensive quality analyzing and hierarchical treatment,the standardized level of improving the fruit and adopting the after treatment,it is a problem urgently to be solved in the present fruit production.With the rapid development of computer technology and the increasingly deepening research of chemo-metrics method,in recent years,as a detection method,near infrared spectrum formation of image nondestructive technology cause the extensive concern,because it has the fast speed of measuring,advantages in many aspects of low grade of cost with high efficiency,apply to relevant trade fields,such as medicine,chemical industry,agriculture,food,etc.more and more.At present,because a sex of inequality that the implicit thing is distributed in fruit,and discrimination model difference and instability that the data processing method of spectral analysis bring,require that the fruit nondestructive detection method based on near infrared spectrum technological fast analysis can be optimized further.Using grapes as the test material,this research try to adopt the method that near infrared spectrum technology combines with chemo-metrics,to set up the method fruit nondestructive measuring and fast quality appraising.First of all we choose the samples used to set up model,and divided into modeling set and verifying set according to certain proportion;Use near infrared spectrum technology to gather and support the primitive spectrum data of a product of samples.The method of adopt the plural scattering correcting(MSC),derivative,level and smoothing etc.carries on the pretreatment to the primitive spectrum,choose the suitable range of wavelength of spectrum according to the content of component part and spectrum characteristic examined;And then utilize partial minimum two(partial least square,PLS)law and differentiate the law of analyzing(discriminant analysis,DA)further,the quantitative model which sets up the near infrared spectrum of fruit and model of determining the nature,used in the fruit sample the content inside including thing is predicted,and the differentiation and qualification of different varieties of grape,maturity,disease worm's fruit.Can't harm measuring and appraising the method with quality to what has been set up and eaten grape fruit raw fast in this experimental study,improve the complexion after treatment level of the fruit and commercialized degree,promote the grape culture production and industry development,have a very important meaning.Have made the following main result in this thesis:1.Using the partial least squares(PLS)to establish a quantitative analysis of each component content of grape quality hybrid index model,and systematically discusses the different spectra pretreatment method,the spectral range of the selected effect the performance of the model prediction.Mixed indicator model refers to,the use of a variety of multiple indexes mixed model is established.Hybrid model can be one-time forecast indicators,the applicable scope is wide.This study established a giant peak summer black,wenk,the disease of Ju feng three mixed quantitative model,analysis to predict the total phenol,total sugar,fructose,sucrose,soluble solids five indicators.Three mixed indicator model adopts multiple scatter correction(MSC)+ first derivative(1 st derivative)+ Norris smoothing pretreatment method,prediction results in addition to a handful of low correlation coefficient,between 0.77 ~ 0.89,the rest are all above 0.90,root mean square error between 0.022 ~ 1.41,characterization model predicted results are good.2.This research has also established a large number of high precision and strong specificity.High precision,strong specificity,higher level as a whole.TQ model built,can use the model for quantitative analysis of unknown components in the sample.Wen ke for example,this study adopt the way of external validation of random wenk spectrum ofunknown components in the sample,the unknown sample spectra import wenk contains five component content of hybrid model forecast;As the diease of Ju feng,for example,will be 10 giant peak second CiGuo sample spectra of unknown components in bulk import two giant peak Ci Guo mixture model to forecast.Has achieved ideal forecast result.3.The research use TQ Analyst in 9 software discriminant analysis(DA)method for Xia hei,Ju meigui,Zui jinxiang,Teng ren,Bi angkou,Jing ya,Hong fushi,Jin shouzhi,Ju feng,Wen ke 10 different varieties of a total of 114 samples of grapes varieties of qualitative identification model is set up.Model after a lot of optimization,the results show that the SNV + first derivative + Norris smoothing or log10 SNV + + Norris smoothing the two groups of pretreatment method,the model of recognition correct rate is highest,reaching 92.11%.In TQ Analyst 9 software,using the principal component scores discriminant model to observe the sample spectrum distribution and the abnormal spectrum,it is concluded that the principal component score 3D figure.In samples of different varieties of grapes in the three dimensional principal component space distribution,can visually see 10 on the spatial distribution of the grape varieties have a relatively clear boundaries,have apparent clustering phenomenon between different varieties of grapes.4.In order to identify the same grape variety of different maturity,the discriminant analysis method of qualitative analysis by TQ 9 Software is operated by using pretreatment methods SNV 1st derivative Norris smoothing and applying qualitative identification of varieties to set up the qualitative model of the three stage in immature,veraison,ripe of Wen ke for a total of 45 grape varieties samples,the recognition rate of the model,was88.89%,the maturity showed obvious distinction.5.In order to identify the different maturity the same grape varieties and disease,TQ Analyst discriminant analysis qualitative analysis software in 9 were used to assess four stages of diseases,immature,veraison,ripe of Jufeng for a total of 52 grape samples to establish the qualitative identification model of product,pretreatment methods MSC is a derivative of Norris smoothing,the correct recognition rate of 96.15%models.According to principal component scores after excluding outliers,the maturity of the samples was clearly different,it was also distinctively different between the disease and healthy fruit. |
PDF Full Text Request