Determination of water content in de-Enzyming green tea leaves based on hyper-Spectral imaging

作者：Yong Guang Hu^*; Pei Pei Chen; Ping Ping Li

来源：Information Technology Journal, 2013, 12(22): 6729-6734.

DOI：10.3923/itj.2013.6729.6734

• Back propagation neural networks
• Gray level co-occurrence matrix
• Hyperspectral Imaging
• PCA
• Principal component regression
• Root-mean-square error of predictions
• Support vector machine regressions
• SVM

To determine water content in de-enzyming green tea leaves rapidly and nondestructively, a prediction model is established based on hyper-spectral imaging technology. Diffuse reflection spectra of 192 samples are collected with hyper-spectral imaging system, among which 144 samples are partitioned to the calibration set and 48 samples are partitioned to the prediction set with a partitioning algorithm based on joint X-Y distance. 3 optimal characteristic wavelengths of 980, 1190 and 1389 nm are selected through principal component analysis. After the preprocessing of image cropping, median filtering and normalization, the eigenvalues of the gray levels and textures are extracted based on gray level co-occurrence matrix. Calibration models of water content are established with principal component regression, back propagation neural network and support vector machine regression based on the eigenvalues above. The results show that the support vector machine regression model with 11 variables is the best and its prediction correlation coefficient is 0.8566 while the root mean square error of prediction is 0.0401. The study provides a rapid and nondestructive way to detect water content in de-enzyming green tea leaves which could be used for online monitoring and feedback control of green tea de-enzyming.

• 出版日期2013
• 单位南京林业大学