Identification of white degradable and non-degradable plastics in food field: A dynamic residual network coupled with hyperspectral technology

In the food field, with the improvement of people's health and environmental protection awareness, degradable plastics have become a trend to replace non-degradable plastics. However, their appearance is very similar, making it difficult to distinguish them. This work proposed a rapid identification method for white non-degradable and degradable plastics. Firstly, a hyperspectral imaging system was used to collect the hyperspectral images of the plastics in visible and near-infrared bands (380-1038 nm). Secondly, a residual network (ResNet) was designed according to the characteristics of hyperspectral information. Finally, a dynamic convolution module was introduced into the ResNet to establish a dynamic residual network (Dy-ResNet) to adaptively mine the data features and realize the classification of the degradable and non-degradable plastics. Dy-ResNet had better classification performance than the other classical deep learning methods. The classification accuracy of the degradable and non-degradable plastics was 99.06%. In conclusion, hyperspectral imaging technology was combined with Dy-ResNet to identify the white non-degradable and degradable plastics effectively.

Applications of Hyperspectral Imaging Technology Combined with Machine Learning in Quality Control of Traditional Chinese Medicine from the Perspective of Artificial Intelligence: A Review

Traditional Chinese medicine (TCM) is the treasure of China, and the quality control of TCM is of crucial importance. In recent years, with the quick rise of artificial intelligence (AI) and the rapid development of hyperspectral imaging (HSI) technology, the combination of the two has been widely used in the quality evaluation of TCM. Machine learning (ML) is the core wisdom of AI, and its progress in rapid analysis and higher accuracy improves the potential of applying HSI to the field of TCM. This article reviewed five aspects of ML applied to hyperspectral data analysis of TCM: partition of data set, data preprocessing, data dimension reduction, qualitative or quantitative models, and model performance measurement. The different algorithms proposed by researchers for quality assessment of TCM were also compared. Finally, the challenges in the analysis of hyperspectral images for TCM were summarized, and the future works were prospected.

Detection of early bruises on loquat using hyperspectral imaging technology coupled with band ratio and improved Otsu method

The bruising is one of the major factors affecting the quality of loquat and the bruised areas of loquat are also prone to harbor bacteria and molds. Therefore, it is critical to detect early bruises of loquat. In this study, a method based on hyperspectral imaging technology coupled with band ratio and improved Otsu method was proposed to detect early bruises of loquat. Firstly, the principal component cluster analysis was used to analyze the three regions of Vis-NIR (397.5-1014.0 nm), Vis (397.5-780.0 nm), and NIR (780.0-1014.0 nm), respectively. It was found that the Vis-NIR and NIR spectral regions along PC1 could be used to effectively distinguish bruised tissues. Then, the key wavelength images corresponding to the two regions were selected according to the load curve, respectively, and two sets of PC images and band ratio images of them were established. After comparison, it was found that the band ratio image Q_{651.3 / 904.3} was the most suitable for subsequent analysis of detecting early bruises of loquat. Finally, in order to evaluate the segmentation effect of the improved Otsu method, the segmentation results of the global threshold and the Otsu method were compared with it, respectively, and it was found that the performance of the improved Otsu method was best. However, since the stem-end area and the bruised area have similar intensity features causing mis-segmentation, the stem-end area was removed by curvature-assisted Hough transform circle detection (CACD) algorithm. And all test set samples were used to evaluate the performance of the proposed method, and the overall accuracy of it was 96.0 %. The results show that the detection method proposed in this study has the potential to detect early bruises of loquat in online practical applications, and it provides a theoretical basis for hyperspectral imaging in the bruise detection of fruit.

Near-infrared hyperspectral imaging technology combined with deep convolutional generative adversarial network to predict oil content of single maize kernel

Rapidly and non-destructively predicting the oil content of single maize kernel is crucial for food industry. However, obtaining a large number of oil content reference values of maize kernels is time-consuming and expensive, and the limited data set also leads to low generalization ability of the model. Here, hyperspectral imaging technology and deep convolutional generative adversarial network (DCGAN) were combined to predict the oil content of single maize kernel. DCGAN was used to simultaneously expand their spectral data and oil content data. After many iterations, fake data that was very similar to the experimental data was generated. Partial least squares regression (PLSR) and support vector regression (SVR) models were established respectively, and their performance was compared before and after data augmentation. The results showed that this method not only improved the performance of two regression models, but also solved the problem of requiring a large amount of training data.

Application of spectral features for separating homochromatic foreign matter from mixed congee

•

A method that can separate homochromatic FM in mixed congee was proposed.

•

Spectral features of FM and mixed congee were extracted to build recognition model.

•

The SVM model achieved high identification rates (99.17%) for homochromatic FM.

•

The proposed method is better than computer vision in separating homochromatic FM.

Foreign matter (FM) in mixed congee not only reduces the quality of the congee but may also harm consumers. However, the common computer vision methods with poor recognition ability for the homochromatic FM. This study used hyperspectral reflectance images with the pattern recognition model to detect homochromatic FM on the mixed congee surface. First, spectral features corresponding to homochromatic FM and background were extracted from hyperspectral images. Then, based on the optimal spectral preprocessing method, LDA, K-nearest neighbor, backpropagation artificial neural network, and support vector machine (SVM) were used to classify the spectral features. The results revealed that the SVM model input with raw spectra principal components exhibited optimal identification rates of 99.17%. Finally, most of the pixels for homochromatic FM were classified correctly by using the SVM model. To summarized, hyperspectral images combined with pattern recognition are an effective method for recognizing homochromatic FM in mixed congee.

Determination of hardness for maize kernels based on hyperspectral imaging

In order to realize rapid and non-destructive detection of hardness for maize kernels, a method for quantitative hardness measurement was proposed based on hyperspectral imaging technology. Firstly, the regression model of hardness and moisture content was established. Then, based on reflectance hyperspectral imaging at wavelengths within 399.75-1005.80 nm, the prediction model of the moisture content was studied by the partial least squares regression (PLSR) based on the characteristic wavelengths, which was selected through successive projection algorithm (SPA). Finally, the hardness prediction model was validated by combing the prediction model of moisture content with the regression model of hardness. The coefficient of determination (R²), the root mean square error (RMSE) the ratio of performance-to-deviation (RPD) and the ratio of error range (RER) of hardness prediction were 0.912, 17.76 MPa, 3.41 and 14, respectively. Therefore, this study provided a method for rapid and non-destructive detection of hardness of maize kernels.

Hyperspectral imaging technology combined with deep forest model to identify frost-damaged rice seeds

In recent years, deep learning models have been widely used in the field of hyperspectral imaging. However, the training of deep learning models requires not only a large number of samples, but also the need to set too many hyper-parameters, which is time consuming and laborious for researchers. This study used hyperspectral imaging technology combined with a deep learning model suitable for small-scale sample data sets, deep forests (DF) model, to classify rice seeds with different degrees of frost damage. During the period, three spectral preprocessing methods (Savitzky-Golay first derivative (SG1), standard normal variate (SNV), and multivariate scatter correction (MSC)) were used to process the original spectral data, and three feature extraction algorithms (principal component analysis (PCA), successive projections algorithm (SPA), and neighborhood component analysis (NCA)) were used to extract the characteristic wavelengths. Moreover, DF model and three traditional machine learning models (decision tree (DT), k-nearest neighbor (KNN), and support vector machine (SVM)) were built based on different numbers of sample sets. After multivariate data analysis, it showed that the pretreatment effect of MSC was the most excellent, and the characteristic wavelength extracted by NCA algorithm was the most useful. In addition, the performance of DF model was better than these three traditional classifier models, and it still performed well in small-scale sample set data. Therefore, DF model was chosen as the best classification model. The results of this study show that the DF model maintains good classification performance in small-scale sample set data, and it has a good application prospect in hyperspectral imaging technology.