Feature construction methods for processing and analysing spectral images and their applications in food quality inspection

Abstract

BackgroundHyperspectral imaging (HSI) technology fusing spectroscopic technology and imaging technology has been proposed to achieve rapid and non-destructive inspection of food quality. In order to make full use of the hyperspectral data containing rich information, it is needed to develop effective and efficient data analysis methods to mine new information from hyperspectral data. The feature construction (FC) method, as a method of extracting information to construct new features, is applied to construct more representative and informative features from hyperspectral data for the detection of food quality. Scope and approachThe review focuses on the construction methods of different dimensional features including zero-dimensional (0-D) features, one-dimensional (1-D) features, two-dimensional (2-D) features and three-dimensional (3-D) features in detail and presents their principles and implementation steps. In the review, applications of the HSI technology combined with different dimensional FC methods for the quality inspection of food are also discussed, and challenges and future work of the HSI technology combined with FC methods are presented. Key findings and conclusionsThe current review is expected to provide some guidance to researchers on different dimensional FC methods, which should further encourage more applications of the HSI technology in food quality inspection. Despite the challenges in combining FC methods with hyperspectral technology, the encapsulation of FC methods in miniaturised hyperspectral instruments is an inevitable trend for future development.