研究目的
To develop a new algorithm for feature variable selection and investigate the feasibility of NIR and MIR methods for potato hierarchical clustering and doneness degree determination.
研究成果
PCFIA is an effective method for feature variable selection, and both NIR and MIR spectroscopic techniques are capable of classifying potato varieties and determining potato doneness degree. The models using selected feature variables performed comparably to full-wavelength methods, demonstrating potential for rapid analysis in food quality control.
研究不足
The robustness of the feature selection has not been confirmed. The variable selection technique should be benchmarked against state-of-the-art methods and tested on other foods with multi-varieties and multi-origins. More multivariate analysis methods (e.g., SIMCA, MLR) should be used to verify efficiency. Further work is needed for mobile device applications.
1：Experimental Design and Method Selection:
The study used NIR and MIR spectroscopy combined with chemometric methods (HCA, PLSR, and a new PCFIA algorithm) for classification and prediction. Samples were prepared by cooking potato tubers to different doneness degrees, and spectra were collected using hyperspectral imaging and microspectroscopic systems.
2：Sample Selection and Data Sources:
Two potato varieties (Rooster and Melody) with 120 samples each from the UK were used. Samples were divided into groups for NIR and MIR experiments, cooked to specific doneness degrees, and spectra were acquired.
3：List of Experimental Equipment and Materials:
Equipment included a pushbroom NIR hyperspectral imaging system (with components like spectrograph, camera, stepper motor), a LUMOS FT-MIR microscope in ATR mode, lab-scale ovens, and software (Matlab, PLS-toolbox, Unscrambler). Materials included potato tubers, ethanol for cleaning, and cotton fabric.
4：Experimental Procedures and Operational Workflow:
Potato samples were peeled, sliced, cooked, and scanned. Spectral data were pre-processed (e.g., first derivative), and multivariate analyses (HCA, PLSR, PCFIA) were performed to classify varieties and predict doneness.
5：Data Analysis Methods:
Data were analyzed using HCA for clustering, PLSR for regression, and PCFIA for feature selection. Statistical measures like correlation coefficients and root mean square errors were used for evaluation.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容-
spectrograph
ImSpector N17E
Spectral Imaging Ltd.
To collect hyperspectral data in the NIR range
-
camera
Xeva 992
Xenics Infrared Solutions
To capture images in the NIR range
-
stepper motor
GPL-DZTSA-1000-X
Zolix Instrument Co.
To control movement in the imaging system
-
FT-MIR microscope
LUMOS
Bruker Optics
To acquire MIR spectral data in ATR mode
-
detector
DTGS
Bruker Optics
To detect MIR signals
-
interferometer
RockSolidTM
Bruker Optics
To perform Fourier transform spectroscopy
-
software
SpectralCube
Spectral Imaging Ltd.
For data acquisition in NIR system
-
software
OPUS 7.2
Bruker Optics
For data acquisition and processing in MIR system
-
illuminating lamps
V-light
Lowel Light Inc.
To provide illumination for imaging
-
ATR crystal
Bruker Optics
To facilitate attenuated total reflectance measurements
-
CCD camera
Bruker Optics
To capture images in the MIR system
-
MCT detector
Bruker Optics
To detect MIR signals with high sensitivity
-
IR beam splitter
Bruker Optics
To split the IR beam for interferometry
-
solid-state laser
Bruker Optics
To provide a reference for the interferometer
-
software
Matlab R2016a
The Mathworks Inc.
For data extraction and calibration
-
software
PLS-toolbox v8.6
Eigenvector Research
For multivariate analysis
-
software
Unscrambler 10.1
Camo Software AS.
For multivariate analysis
-
software
Matlab R2017b
The Mathworks Inc.
For multivariate analysis
-
oven
To cook potato samples for NIR experiment
-
microwave oven
To cook potato samples for MIR experiment
-
登录查看剩余18件设备及参数对照表
查看全部
