Principal component discriminant analysis for feature extraction and classification of hyperspectral images

In this paper, a novel feature extraction technique called SuperMNF is proposed, which is an extension of the minimum noise fraction (MNF) transformation. In SuperMNF, each superpixel has its own transformation matrix and MNF transformation is performed on each superpixel individually. The basic idea behind the SuperMNF is that each superpixel contains its specific signal and noise covariance matrices which are different from the adjacent superpixels. The extracted features, owning spatial-spectral content and provided in the lower dimension, are classified by maximum likelihood classifier and support vector machines. Experiments that are conducted on two real hyperspectral images, named Indian Pines and Pavia University, demonstrate the efficiency of SuperMNF since it yielded more promising results than some other feature extraction methods (MNF, PCA, SuperPCA, KPCA, and MMP).

Download Full-text

Identification of mechanical damage in the 'Fuji' apple cv. using artificial hyperspectral vision

DYNA ◽

10.15446/dyna.v86n210.78605 ◽

2019 ◽

Vol 86 (210) ◽

pp. 224-232

Author(s):

Oscar Leonardo García Navarrete ◽

Sergio Cubero García ◽

José Manuel Prats Montalbán

Keyword(s):

Discriminant Analysis ◽

Impact Damage ◽

Mechanical Damage ◽

Hyperspectral Images ◽

Different Temperatures ◽

Fuji Apple ◽

First Time ◽

Sampling Times ◽

Over Time

One problem in the post-harvest phase of apples is the mechanical impact damage; its identification prevents quality issues during storage. The objective was to identify the wavelengths at which the damage is detected early in apples of the 'Fuji' cultivar, simulating the damage with a controlled stroke and taking hyperspectral images from 400 to 1700 nm. Three experiments were carried out at different temperatures (4 and 20 ° C) and with varying sampling times. It was found that in the NIR zone ranging between 1050 and 1100 nm, it was possible to classify healthy and bruised zones by means of a discriminant analysis by partial least squares (PLS-DA). Additionally, the evolution of the damage over time was not significant for the classification of the pixels (healthy and bruised classes), since bumps were detected in all three experiments from the first time.

Download Full-text

Determination of HPLC-UV Fingerprints of Spanish Paprika (Capsicum annuum L.) for Its Classification by Linear Discriminant Analysis

Sensors ◽

10.3390/s18124479 ◽

2018 ◽

Vol 18 (12) ◽

pp. 4479 ◽

Cited By ~ 6

Author(s):

Xavier Cetó ◽

Núria Serrano ◽

Miriam Aragó ◽

Alejandro Gámez ◽

Miquel Esteban ◽

...

Keyword(s):

Discriminant Analysis ◽

Linear Discriminant Analysis ◽

Principal Component ◽

Reversed Phase ◽

Classification Rate ◽

Phenolic Profile ◽

Linear Discriminant ◽

Sample Extraction

The development of a simple HPLC-UV method towards the evaluation of Spanish paprika’s phenolic profile and their discrimination based on the former is reported herein. The approach is based on C18 reversed-phase chromatography to generate characteristic fingerprints, in combination with linear discriminant analysis (LDA) to achieve their classification. To this aim, chromatographic conditions were optimized so as to achieve the separation of major phenolic compounds already identified in paprika. Paprika samples were subjected to a sample extraction stage by sonication and centrifugation; extracting procedure and conditions were optimized to maximize the generation of enough discriminant fingerprints. Finally, chromatograms were baseline corrected, compressed employing fast Fourier transform (FFT), and then analyzed by means of principal component analysis (PCA) and LDA to carry out the classification of paprika samples. Under the developed procedure, a total of 96 paprika samples were analyzed, achieving a classification rate of 100% for the test subset (n = 25).

Download Full-text