Deep learning classifiers for near infrared spectral imaging: a tutorial

Abstract

Deep learning (DL) has recently achieved considerable successes in a wide range of applications, such as speech recognition, machine translation and visual recognition. This tutorial provides guidelines and useful strategies to apply DL techniques to address pixel-wise classification of spectral images. A one-dimensional convolutional neural network (1-D CNN) is used to extract features from the spectral domain, which are subsequently used for classification. In contrast to conventional classification methods for spectral images that examine primarily the spectral context, a three-dimensional (3-D) CNN is applied to simultaneously extract spatial and spectral features to enhance classificationaccuracy. This tutorial paper explains, in a stepwise manner, how to develop 1-D CNN and 3-D CNN models to discriminate spectral imaging data in a food authenticity context. The example image data provided consists of three varieties of puffed cereals imaged in the NIR range (943–1643 nm). The tutorial is presented in the MATLAB environment and scripts and dataset used are provided. Starting from spectral image pre-processing (background removal and spectral pre-treatment), the typical steps encountered in development of CNN models are presented. The example dataset provided demonstrates that deep learning approaches can increase classification accuracy compared to conventional approaches, increasing the accuracy of the model tested on an independent image from 92.33 % using partial least squares-discriminant analysis to 99.4 % using 3-CNN model at pixel level. The paper concludes with a discussion on the challenges and suggestions in the application of DL techniques for spectral image classification.

Deep Coding Patterns Design for Compressive Near-Infrared Spectral Classification

“…The classification network structure used in the E2E model was inspired in the work presented in [22] for NIR images. Specifically, the network consists of three convolutional layers with batch normalization and rectified linear unit (ReLU) as the activation function, followed by a dropout to avoid overfitting, and finally, two fully connected layers, where the last one has the softmax activation.…”

“…The performance of the proposed approach is demonstrated on two hyperspectral datasets in the NIR region: Cereals [21] in the spectral range of 943-1643 nm, and Yatsuhashi, which covers 1293-2215 nm. The results show better average classification accuracy above 90% for different sensing ratios compared to coding strategies based on the visible range [15] and [22], specifically, when the sensing ratio decreases, where the reached difference is up to 10% of accuracy.…”

Identification of micro- and nanoplastics released from medical masks using hyperspectral imaging and deep learning

“…It is assumed that the preprocessing step allowed the reduction of the spectral noise, permitting more accurate edge segmentation. 68 The calculated accuracy metrics revealed that the averaged performance of SAM and CNN is on the same level, with the highest F-score in the 1D-CNN SNV model and the highest accuracy in SAM (Table 1 and Table S1 †). It is also noteworthy that the normalisation of spectra could both decrease the performance of CNN, as shown for the model built with MSC-corrected data, and increase classification accuracy as in the case of applying the SNV-corrected dataset.…”

“…31,69 Various works in using CNNs for hyperspectral data analysis addressed the problem of extracting and preprocessing reflectance from hyperspectral images. 68,70,71 Typically, the SNV correction stays as an optimal method for preserving the general trend of spectra with unwanted noise reduction, which enhances the classification of neural networks. At the same time, modifications and combinations of traditional preprocessing steps, depending on the material under investigation and spectroscopy conditions, may be used to improve the performance of different models.…”

Discriminating the Viscoelastic Properties of Cellulose Textile Fibers for Recycling

“…This indicated underand over-fitting and emphasized the importance of extracting features for LDA with our dataset. We also determined cross-validation accuracy by using different preprocessing combinations based on first and second derivative Savitzky-Golay filtering and SNV (Xu et al, 2020). The results are shown in Fig.…”