Abstract

Near-infrared hyperspectral imaging technology was adopted in this study to discriminate among varieties of raisins produced in Xinjiang Uygur Autonomous Region, China. Eight varieties of raisins were used in the research, and the wavelengths of the hyperspectral images were from 900 to 1700 nm. A novel waveform resolution method is proposed to reduce the hyperspectral data and extract the features. The waveform-resolution method compresses the original hyperspectral data for one pixel into five amplitudes, five frequencies, and five phases for 15 feature values in all. A neural network was established with three layers—eight neurons for the first layer, three neurons for the hidden layer, and one neuron for the output layer—based on the 15 features used to determine the varieties of raisins. The accuracies of the model, which are presented as sensitivity, precision, and specificity, for the testing data set, are 93.38, 81.92, and 99.06%. This is higher than the accuracies of the model using a conventional principal component analysis feature-extracting method combined with a neural network, which has a sensitivity of 82.13%, precision of 82.22%, and specificity of 97.45%. The results indicate that the proposed waveform-resolution feature-extracting method combined with hyperspectral imaging technology is an efficient method for determining varieties of raisins.

Feature extraction using Mel frequency cepstral coefficients for hyperspectral image classification

Delian Liu, Xiaorui Wang, Jianqi Zhang, and Xi Huang
Appl. Opt. 49(14) 2670-2675 (2010)

Wavelet based feature extraction and visualization in hyperspectral tissue characterization

Martin Denstedt, Asgeir Bjorgan, Matija Milanič, and Lise Lyngsnes Randeberg
Biomed. Opt. Express 5(12) 4260-4280 (2014)

Feature weighting algorithms for classification of hyperspectral images using a support vector machine

Bin Qi, Chunhui Zhao, and Guisheng Yin
Appl. Opt. 53(13) 2839-2846 (2014)

Hyperspectral image super-resolution based on the transfer of both spectra and multi-level features

Xuheng Cao, Yusheng Lian, Zilong Liu, Han Zhou, Xiangmei Hu, Beiqing Huang, and Wan Zhang
Opt. Lett. 47(14) 3431-3434 (2022)

Tumor tissue classification based on micro-hyperspectral technology and deep learning

Bingliang Hu, Jian Du, Zhoufeng Zhang, and Quan Wang
Biomed. Opt. Express 10(12) 6370-6389 (2019)