Raman spectroscopy is a popular technology for material identification, but it encounters difficulties when distinguishing components with closely related substances by intuition and experience. The distinction of fats and oils is a typical example, which is significant in the food industry. In this work, the Raman spectroscopy and deep learning algorithm are combined to analyze closely related animal fats (lard, butter, mutton fat and chicken fat) and vegetable oils (soybean oil and peanut oil) in a dataset. A deep neural network founded on the VGG architecture with attention mechanism is developed, reaching an accuracy of 100% for fats and oils classification. By combining Raman spectroscopy with deep learning, this research provides a potent technique for tackling the identification of similar substances.

The identification of mixed solutions is a challenging and important subject in chemical analysis. In this paper, we propose a novel workflow that enables rapid qualitative and quantitative detection of mixed solutions. We use a methanol-ethanol mixed solution as an example to demonstrate the superiority of this workflow. The workflow includes the following steps: (1) converting Raman spectra into Raman images through CWT; (2) using MobileNetV3 as the backbone network, improved multi-label and multi-channel synchronization enables simultaneous prediction of multiple mixture concentrations; and (3) using transfer learning and multi-stage training strategies for training to achieve accurate quantitative analysis. We compare six traditional machine learning algorithms and two deep learning models to evaluate the performance of our new method. The experimental results show that our model has achieved good prediction results when predicting the concentration of methanol and ethanol, and the coefficient of determination R2 is greater than 0.999. At different concentrations, both MAPE and RSD outperform other models, which demonstrates that our workflow has outstanding analytical capabilities. Importantly, we have solved the problem that current quantitative analysis algorithms for Raman spectroscopy are almost unable to accurately predict the concentration of multiple substances simultaneously. In conclusion, it is foreseeable that this non-destructive, automated, and highly accurate workflow can further advance Raman spectroscopy.

Raman spectroscopy is a general and non-destructive detection technique that can obtain detailed information of the chemical structure of materials. In the past, when using chemometric algorithms to analyze the Raman spectra of mixtures, the challenges of complex spectral overlap and noise often limited the accurate identification of components. The emergence of deep learning has introduced a novel approach to qualitative analysis of mixed Raman spectra. In this paper, we propose a deep learning-based Raman spectroscopy qualitative analysis algorithm (RST) by borrowing the ideas of convolutional neural network and Transformer. By transforming the Raman spectrum into 64 word vectors, the contribution weights of each word vector to the components are obtained. For the 75 spectral data used for validation, the positive identification rate can reach 100.00%, the recall rate can reach 99.3%, the average identification score can reach 9.51, and it is applicable to the fields of Raman and surface-enhanced Raman spectroscopy. Furthermore, compared with traditional CNN models, RST has excellent accuracy and robustness in identifying components in complex mixtures. The model's interpretability has been enhanced, aiding in a deeper understanding of spectroscopic learning patterns for future analysis of more complex mixtures.

Enhancing the quality of spectral denoising plays a vital role in Raman spectroscopy. Nevertheless, the intricate nature of the noise, coupled with the existence of impurity peaks, poses significant challenges to achieving high accuracy while accommodating various Raman spectral types. In this study, an innovative adaptive sparse decomposition denoising (ASDD) method is proposed for denoising Raman spectra. This approach features several innovations. Firstly, a dictionary comprising spectral feature peaks is established from the input spectra by applying a chemometric feature extraction method, which better aligns with the original data compared to traditional dictionaries. Secondly, a dynamic Raman spectral dictionary construction technique is introduced to swiftly adapt to new substances, employing a limited amount of additional Raman spectral data. Thirdly, the orthogonal matching pursuit algorithm is utilized to sparsely decompose the Raman spectra onto the constructed dictionaries, effectively eliminating various random and background noises in the Raman spectra. Empirical results confirm that ASDD enhances the accuracy and robustness of denoising Raman spectra. Significantly, ASDD surpasses existing algorithms in processing Raman spectra of pesticide.

The disparity in hardware quality among various models of Raman spectrometers gives rise to variations in the acquired Raman spectral data, even when the same substance is collected under identical external conditions. Conventionally, models constructed using data obtained from a particular instrument exhibit issues such as limited applicability or poor performance when deployed to different instruments. Currently, numerous model transfer algorithms grounded in chemometrics have been developed, all aiming to establish a mapping relationship capable of transforming spectral data from the source domain to the target domain. With the advancement of deep learning techniques, the utilization of deep learning enables the effective resolution of nonlinear mapping relationships between two spectral vectors. In the field of image translation, the Cycle-Consistent Adversarial Networks, Cycle-GAN, has already achieved mutual transformation between two distinct style images. However, due to images being multidimensional matrix data, unlike one-dimensional spectral data vectors, we have constructed a deep learning network based on Cycle-GAN for vector-to-vector transformation. This network allows the direct conversion of spectral data from the source domain to the target domain, without requiring parameter adjustments or other operations. Compared with traditional chemometric methods, our method is more intelligent and efficient. Finally, the cosine similarity between the source domain data and the transformed target domain data exceeds 99%.

In the context of the ongoing COVID-19 outbreak, the regulation of the use of personal protective items, specifically helmets and masks, on construction sites is of critical importance. Due to the complex nature of construction environments and the high number of workers, relying on worker inspection and surveillance cameras to detect the wearing of helmets and masks has problems such as poor timeliness, low accuracy, and low efficiency. This paper provides a deep learning approach to address the above issues. We employed YOLOv5s to train separate models for helmet and mask detection, achieving accuracy and mAP@0.5 of greater than 90% for both models. Tested on the divided test set, the mAP@0.5 for both helmet and mask detection were over 96%, demonstrating the effectiveness of our models. These models can be deployed in real environments, effectively solving the problem of monitoring the compliance of construction workers wearing personal protective equipment.

Raman spectroscopy is widely used in the identification of substances. Raman spectra contain molecular information from various components and interference from noise and instruments. Therefore, using Raman spectroscopy to identify components is still challenging, especially for substances with high similarity. In this study, a class of highly similar products (methanol and ethanol) and a multicomponent mixture of propanol and water were tested using a portable Raman spectrometer. They are divided into 11 categories according to the volume fraction ratio. A total of 5,060 groups of Raman spectrum data were obtained, constituting the data set of this study. The deep neural network structure adopted in this study is based on ResNet architecture, on which the SE module in the attention mechanism is added, which increases the weight of some spectral features and achieves significant performance improvement. Finally, SE-ResNet achieves a recall of 0.95, a precision rate of 0.95, a Micro-F1 of 0.95, and an accuracy rate of 99.20%.

Raman spectroscopy detection technology is a modern detection technology that combines chemometrics and is widely used in many fields. The application range includes biomedicine, food safety, medical diagnosis, oil detection, etc. It has the advantages of high speed, accuracy, flexible application and high sensitivity. With excellent performance in feature extraction and image recognition, deep learning has attracted more attention in the application field of Raman spectroscopy. The Raman spectroscopy detection technology combined with deep learning method can be optimized in terms of accuracy, detection speed and other performance indicators. This paper reviewed the application and progress of Raman spectroscopy in substance detection and various industries, reviewed the analysis processes and processing methods of Raman spectroscopy methods, including traditional chemometric methods such as multiple linear regression (MLR), principal component analysis (PCA), linear discriminant analysis (LDA), partial least squares (PLS), etc., and spectral signal preprocessing and deep learning analysis methods. In the end, the advantages and disadvantages of Raman spectroscopy detection technology based on deep learning and its development prospects were discussed.

In recent years, deep learning has been widely used in the field of Raman spectral classification. However, the majority of the training and test sets are generated by the same device (generally a portable Raman spectrometer), with little difference between them, and the trained model may not be directly applicable to other devices. In this study, we established a database of six cephalosporin Raman spectra and proposed a classification algorithm VGGNeXt for cephalosporin Raman spectra. VGGNeXt takes inspiration from ConvNeXt, borrows some tricks from Swin-T, and re-improves VGG. Training data were high-resolution spectra from a benchtop Raman spectrometer, and test data were low-resolution spectra from a portable Raman spectrometer. The impact of preprocessing and dataset size on algorithm accuracy was explored. The results show that our network outperforms other comparative algorithms in all cases. After preprocessing, the VGGNeXt model achieves 100% accuracy on both full and halved data sets, and 99.9% accuracy when there are only 10 data for each cephalosporin class. The results show that the experimental ideas and processing methods in this paper solve the problems of model transfer and instrument standardization to a certain extent, and the model has good robustness.