The Detection Method Research On Delinted Cottonseeds' Vigor Based On Hyperspectral Imaging

cottonseed is the basis of cottonseed production,and the quality of cottonseed directly affect the cotton yield and fiber quality.At present,the method of detecting the vigor of delinted cottonseed is that the detection time is too long,the labor intensity is large and the efficiency is low.In this paper,hyperspectral image technique was used to non-destructively detect delinted cottonseed.The three varieties of delinted cottonseed,such as Xin Luzao50,Xin Luzao57,Xin Luzao62 were used as the research object.The germination length measured by the conductivity and low temperature germination test of delintedt cottonseed was the main index.Not only do achieve delinted cottonseed vigor non-destructive,rapid detection,but also laid the theoretical basis for the scientific and accurate assessment of the vigor of delinted cottonseed.(1)Make a qualitative analysis of detecting delinted cottonseed vigor.In this study,the hyperspectral images of the three varieties of delinted cottonsees were collected for 24 and 48 hours by artificial accelerated aging.The hyperspectral images were obtained.The data of delinted cottonseed vigor were obtained after conductivity and low temperature germination experiment.To dimensionality reduction,the spectral data of the delinted cottonseed with no aging,artificial accelerated aging for 24 hours and artificial accelerated aging 48 hours were analyzed by principal component analysis.The method of discriminant analysis and support vector machine pattern recognition was used to analyze.The K-CV and Particle Swarm Optimization were used to optimize the Support Vector Machine parameters.The first 10 principal components are used as input variables to establish the discriminant analysis and support vector machine model.The comparison with discriminant analysis and support vector machine classification accuracy,the discriminant analysis model is better,the classification accuracy can reach 89.70%.(2)Make a quantitative analysis of conductivity of delinted cottonseed.The results of spectral pretreatment were compared and analyzed.Finally,SNV + first order differential +S-G smoothing and SNV + first order differential + norris were used to deal with the best.Three varieties of velvet cotton varieties using S-G smooth and Norris smooth the same number of points used,the number of different.The Chauvenet and Leverage methods were used to compare the abnormal samples.The PLS,SMLR and PCR models were established for the conductivity of velvet cotton.The results of the model were the PLS model,RMSEP and RMSEC were 38 ~ 46 ?S / cm.The gap of RMSEP and RMSEC was small,the Rc and pRhad reached more than 0.88.The optimal set of characteristic wavelengths is extracted by siPLS-GA,and the Rc andpRof the model were 0.90,0.89.RMSEC and RMSEP were 48.20?S / cm and 42.60 ?S / cm,respectively.(3)Make a quantitative analysis of length of delinted cottonseed.The dynamic parameters were obtained by low temperature germination test.The PLS,SMLR and PCR models of germination length were established by pretreatment of the spectral data and the removal of abnormal samples.The results show that the PLS model is the best,cR andpRreach 0.60 or more,RMSEC and RMSEP range is 1.60 ~ 1.76 cm.The optimal set of characteristic wavelengths is extracted by siPLS-GA,and the Rc andpRof the model were 0.70,0.65.RMSEC and RMSEP were 1.57 cm and 1.60 cm,respectively.(4)Designed a set of delinted cottonseed vigor detection software based MATAB as the development platform.The functional modules of the system include five parts,namely data import function module,spectral data preprocessing function module,delinted cottonseed modeling sample selection function module,modeling band selection function module and dynamic detection function module.