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

作者：Hongbin Pu, Jingxiao Yu, Da‐Wen Sun, Qingyi Wei, Zhe Wang · 发表于：Trends in Food Science & Technology · 年份：2023 · DOI：10.1016/j.tifs.2023.06.036 · 被引用次数：44 · 研究领域：Spectroscopy and Chemometric Analyses、Remote-Sensing Image Classification、Advanced Chemical Sensor Technologies

Hyperspectral 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. The 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. The 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 instrument...