Using hyperspectral analysis technology combined with pattern recognition

the classification and detection methods of eight potato micro seed potatoes(Daxiyang

Holland-14

Holland fifteen 041

Holland fifteen Q8

Jizhangshu 12

Jizhangshu 8

Xingjia 2 and Y2) were established.276 seed tuber samples were collected.The original spectra of 860-1 700 nm were preprocessed by standardize

11 points Savitzky-Golay smoothing and 4 points differential first derivative.Principal component analysis showed that the cumulative contribution rate of the first three principal components was 95.12%

including most information of the original spectra

and could be used as classification variables.Then

linear discriminant analysis

BP neural network and support vector machine were used for classification modeling.Finally

the classification models of 8 potato micro seed potatos were established by stratification and step by step.Firstly

the linear discriminant analysis model was used to distinguish Daxiyang

Holland-14

Holland fifteen 041 and other varieties.The average correct recognition rate was 88.79%.Then BP neural network model was established to divide the samples of other varieties into two categories:Jizhangshu 8

Y2

and Holland fifteen Q8

Jizhangshu 12

Xingjia 2

with an average correct recognition rate of 93.24%.Finally

the BP neural network model was used to distinguish Jizhangshu 8 and Y2

with the average correct recognition rate of 77.78%;and the support vector machine classification model was used to distinguish Holland fifteen Q8

Jizhangshu 12 and Xingjia 2

with the average correcct recognition rate of 8723%.The method was applied to the classification detection of eight potato seed potatos with the average correct recognition rate of 89.75%

which indicated that the hyperspectral analysis technology could be used for the classification and detection of potato micro seed potatos.