Origins classification of egg with different storage durations using FT-NIR: A characteristic wavelength selection approach based on information entropy. (October 2022)

• Record Type: Journal Article
• Authors: Liu, Chengkang; Wang, Qiaohua; Lin, Weiguo; Yu, Chengdong
• Is Part Of: Biosystems engineering. Volume 222(2022)
• Page Start: 82
• Page End: 92
• Publication Date: 2022-10
• Subjects: Egg freshness -- Preprocessing -- Reflectance spectrum -- Eggshell -- Support vector machine -- Trace; Bioengineering -- Periodicals; Agricultural engineering -- Periodicals; Biological systems -- Periodicals; Génie rural -- Périodiques; Systèmes biologiques -- Périodiques; 631
• DOI: 10.1016/j.biosystemseng.2022.07.016
• Languages: English
• ISSNs: 1537-5110

Abstract

Abstract : The origin of fresh eggs is an essential determinant of their quality. This study's main goal was to classify fresh eggs' origins, after the characteristic wavelength from the near-infrared reflectance spectral of the eggs, using a support vector machine (SVM) for classification. To improve classification accuracy, eight categories of eggs were treated as classification targets. A Fourier transform near-infrared (FT-NIR) spectrometer was used to acquire spectral data of 800 eggs. A characteristic wavelength selection approach was presented based on information entropy (CWSABIE). Genetic algorithm-partial least squares (GA-PLS), interval partial least squares (iPLS), and competitive adaptive reweighting sampling (CARS) were compared. Standard normal variable transformation (SNV), Savitzky–Golay filtering, and Centralisation were applied to preprocess spectral data. The results indicate that when using CWSABIE with SNV, the model had the highest accuracy (93.8%) and can be used to classify the data of eggs after 15 days of storage (91.4%). The model show potential for application to online inspection for eggs from different storage days. Highlights: A characteristic wavelength selection approach based on information entropy was suggested. Standard normal variate transformation and normalisation can reduce random error in experiments to varying degrees. The model can classify the data of eggs stored on different days. The spectrum of eggshells contains geographical information.