Geographical classification of Nanfeng mandarin by near infrared spectroscopy coupled with chemometrics methods

Near infrared spectroscopy(NIRS),coupled with principal component analysis and wavelength selection techniques,has been sed to develop a robust and reliable reduced-spectrum classifi-cation model for determining the geographical origins of Nanfeng mandarins.The application of the changeable size moving window principal component analysis(CSMWPCA)provided a notably improved lassification model,with correct classification rates of 92.00%,100.00%,90.00%,100.00%,100.00%,100.00%and 100.00%for Fujian,Guangxi,Hunan,Baishe,Baofeng,Qiawan,Sanxi samples,respectively,as well as,a total dassification rate of 97.52%in the wavelength range from 1007 to 1296 nm.To test and apply the proposed method,the procedure was applied to the analysis of 59 samples in an independent test set.Good identification results(correct rate of 96.61%)were also received.The improvement achieved by the application of CSMWPCA method was particularly remarkable when taking the low complexities of the final model(290 variables)into account.The results of the study showed the great potential of NIRS as a fast,nondestructive and environmentally acceptable method for the rapid and reliable determination for geographical classifcation of Nanfeng mandarins.

Optimal Feature Extraction Using Greedy Approach for Random Image Components and Subspace Approach in Face Recognition

An innovative and uniform framework based on a combination of Gabor wavelets with principal component analysis （PCA） and multiple discriminant analysis （MDA） is presented in this paper. In this framework, features are extracted from the optimal random image components using greedy approach. These feature vectors are then projected to subspaces for dimensionality reduction which is used for solving linear problems. The design of Gabor filters, PCA and MDA are crucial processes used for facial feature extraction. The FERET, ORL and YALE face databases are used to generate the results. Experiments show that optimal random image component selection （ORICS） plus MDA outperforms ORICS and subspace projection approach such as ORICS plus PCA. Our method achieves 96.25%, 99.44% and 100% recognition accuracy on the FERET, ORL and YALE databases for 30% training respectively. This is a considerably improved performance compared with other standard methodologies described in the literature.