Quality classification of cooked, sliced turkey hams using NIR hyperspectral imaging system

A deep learning model for rapid classification of tea coal disease

BACKGROUND

The common tea tree disease known as "tea coal disease" (Neocapnodium theae Hara) can have a negative impact on tea yield and quality. The majority of conventional approaches for identifying tea coal disease rely on observation with the human naked eye, which is labor- and time-intensive and frequently influenced by subjective factors. The present study developed a deep learning model based on RGB and hyperspectral images for tea coal disease rapid classification.

RESULTS

Both RGB and hyperspectral could be used for classifying tea coal disease. The accuracy of the classification models established by RGB imaging using ResNet18, VGG16, AlexNet, WT-ResNet18, WT-VGG16, and WT-AlexNet was 60%, 58%, 52%, 70%, 64%, and 57%, respectively, and the optimal classification model for RGB was the WT-ResNet18. The accuracy of the classification models established by hyperspectral imaging using UVE-LSTM, CARS-LSTM, NONE-LSTM, UVE-SVM, CARS-SVM, and NONE-SVM was 80%, 95%, 90%, 61%, 77%, and 65%, respectively, and the optimal classification model for hyperspectral was the CARS-LSTM, which was superior to the model based on RGB imaging.

CONCLUSIONS

This study revealed the classification potential of tea coal disease based on RGB and hyperspectral imaging, which can provide an accurate, non-destructive, and efficient classification method for monitoring tea coal disease.

A quantitative on-line analysis of salt in cured ham by near-infrared spectroscopy and chemometrics

In this work, non-invasive near-infrared spectroscopy (NIRS) combined with chemometrics was evaluated as a possible online analytical technique to categorize pieces of cured ham on the industrial production line based on their maximum sodium content. Stifle muscle was selected for the development of the NIRS prediction models because it is the one with the highest sodium content and the easiest in terms of accessibility for spectral measurement. In the study, samples with varying thicknesses were taken. The suitability of this method is demonstrated when a 5 mm sample is used for the construction of the model, obtaining the best fit with an R²cv of 92% and a prediction error of 0.11% sodium that coincides with the error of the reference method. In conclusion, a method is proposed for the direct determination of sodium content on the production line which allows the different pieces of ham to be quickly categorized according to their salt content.

Evaluation of Soybean Wildfire Prediction via Hyperspectral Imaging

Plant diseases that affect crop production and productivity harm both crop quality and quantity. To minimize loss due to disease, early detection is a prerequisite. Recently, different technologies have been developed for plant disease detection. Hyperspectral imaging (HSI) is a nondestructive method for the early detection of crop disease and is based on the spatial and spectral information of images. Regarding plant disease detection, HSI can predict disease-induced biochemical and physical changes in plants. Bacterial infections, such as Pseudomonas syringae pv. tabaci, are among the most common plant diseases in areas of soybean cultivation, and have been implicated in considerably reducing soybean yield. Thus, in this study, we used a new method based on HSI analysis for the early detection of this disease. We performed the leaf spectral reflectance of soybean with the effect of infected bacterial wildfire during the early growth stage. This study aimed to classify the accuracy of the early detection of bacterial wildfire in soybean leaves. Two varieties of soybean were used for the experiment, Cheongja 3-ho and Daechan, as control (noninoculated) and treatment (bacterial wildfire), respectively. Bacterial inoculation was performed 18 days after planting, and the imagery data were collected 24 h following bacterial inoculation. The leaf reflectance signature revealed a significant difference between the diseased and healthy leaves in the green and near-infrared regions. The two-way analysis of variance analysis results obtained using the Python package algorithm revealed that the disease incidence of the two soybean varieties, Daechan and Cheongja 3-ho, could be classified on the second and third day following inoculation, with accuracy values of 97.19% and 95.69%, respectively, thus proving his to be a useful technique for the early detection of the disease. Therefore, creating a wide range of research platforms for the early detection of various diseases using a nondestructive method such HSI is feasible.

Classification of emotional stress and physical stress using a multispectral based deep feature extraction model

A classification model (Stress Classification-Net) of emotional stress and physical stress is proposed, which can extract classification features based on multispectral and tissue blood oxygen saturation (StO2) characteristics. Related features are extracted on this basis, and the learning model with frequency domain and signal amplification is proposed for the first time. Given that multispectral imaging signals are time series data, time series StO2 is extracted from spectral signals. The proper region of interest (ROI) is obtained by a composite criterion, and the ROI source is determined by the universality and robustness of the signal. The frequency-domain signals of ROI are further obtained by wavelet transform. To fully utilize the frequency-domain characteristics, the multi-neighbor vector of locally aggregated descriptors (MN-VLAD) model is proposed to extract useful features. The acquired time series features are finally put into the long short-term memory (LSTM) model to learn the classification characteristics. Through SC-NET model, the classification signals of emotional stress and physical stress are successfully obtained. Experiments show that the classification result is encouraging, and the accuracy of the proposed algorithm is over 90%.

Characterisation and Identification of Individual Intact Goat Muscle Samples (Capra sp.) Using a Portable Near-Infrared Spectrometer and Chemometrics

Adulterated, poor-quality, and unsafe foods, including meat, are still major issues for both the food industry and consumers, which have driven efforts to find alternative technologies to detect these challenges. This study evaluated the use of a portable near-infrared (NIR) instrument, combined with chemometrics, to identify and classify individual-intact fresh goat muscle samples. Fresh goat carcasses (n = 35; 19 to 21.7 Kg LW) from different animals (age, breeds, sex) were used and separated into different commercial cuts. Thus, the longissimus thoracis et lumborum, biceps femoris, semimembranosus, semitendinosus, supraspinatus, and infraspinatus muscles were removed and scanned (900–1600 nm) using a portable NIR instrument. Differences in the NIR spectra of the muscles were observed at wavelengths of around 976 nm, 1180 nm, and 1430 nm, associated with water and fat content (e.g., intramuscular fat). The classification of individual muscle samples was achieved by linear discriminant analysis (LDA) with acceptable accuracies (68–94%) using the second-derivative NIR spectra. The results indicated that NIR spectroscopy could be used to identify individual goat muscles.

Prediction of Tea Polyphenols, Free Amino Acids and Caffeine Content in Tea Leaves during Wilting and Fermentation Using Hyperspectral Imaging

The withering and fermentation degrees are the key parameters to measure the processing technology of black tea. The traditional methods to judge the degree of withering and fermentation are time-consuming and inefficient. Here, a monitoring model of the biochemical components of tea leaves based on hyperspectral imaging technology was established to quantitatively judge the withering and fermentation degrees of fresh tea leaves. Hyperspectral imaging technology was used to obtain the spectral data during the withering and fermentation of the raw materials. The successive projections algorithm (SPA), competitive adaptive reweighted sampling (CARS), and uninformative variable elimination (UVE) are used to select the characteristic bands. Combined with the support vector machine (SVM), random forest (RF), and partial least square (PLS) methods, the monitoring models of the tea polyphenols (TPs), free amino acids (FAA) and caffeine (CAF) contents were established. The results show that: (1) CARS performs the best among the three feature band selection methods, and PLS performs the best among the three machine learning models; (2) the optimal models for predicting the content of the TPs, FAA, and CAF are CARS-PLS, SPA-PLS, and CARS-PLS, respectively, and the coefficient of determination of the prediction set is 0.91, 0.88, and 0.81, respectively; and (3) the best models for quantitatively judging the withering and fermentation degrees are FAA-SPA-PLS and TPs-CARS-PLS, respectively. The model proposed in this study can improve the monitoring efficiency of the biochemical components of tea leaves and provide a basis for the intelligent judgment of the withering and fermentation degrees in the process of black tea processing.

Algorithm for Mapping Kidney Tissue Water Content during Normothermic Machine Perfusion Using Hyperspectral Imaging

The preservation of kidneys using normothermic machine perfusion (NMP) prior to transplantation has the potential for predictive evaluation of organ quality. Investigations concerning the quantitative assessment of physiological tissue parameters and their dependence on organ function lack in this context. In this study, hyperspectral imaging (HSI) in the wavelength range of 500–995 nm was conducted for the determination of tissue water content (TWC) in kidneys. The quantitative relationship between spectral data and the reference TWC values was established by partial least squares regression (PLSR). Different preprocessing methods were applied to investigate their influence on predicting the TWC of kidneys. In the full wavelength range, the best models for absorbance and reflectance spectra provided Rp2 values of 0.968 and 0.963, as well as root-mean-square error of prediction (RMSEP) values of 2.016 and 2.155, respectively. Considering an optimal wavelength range (800–980 nm), the best model based on reflectance spectra (Rp2 value of 0.941, RMSEP value of 3.202). Finally, the visualization of TWC distribution in all pixels of kidneys’ HSI image was implemented. The results show the feasibility of HSI for a non-invasively and accurate TWC prediction in kidneys, which could be used in the future to assess the quality of kidneys during the preservation period.

Wavelength Selection FOR Rapid Identification of Different Particle Size Fractions of Milk Powder Using Hyperspectral Imaging

Hyperspectral imaging (HSI) in the spectral range of 400-1000 nm was tested to differentiate three different particle size fractions of milk powder. Partial least squares discriminant analysis (PLS-DA) was performed to observe the relationship of spectral data and particle size information for various samples of instant milk powder. The PLS-DA model on full wavelengths successfully classified the three fractions of milk powder with a coefficient of prediction 0.943. Principal component analysis (PCA) identified each of the milk powder fractions as separate clusters across the first two principal components (PC₁ and PC₂) and five characteristic wavelengths were recognised by the loading plot of the first three principal components. Weighted regression coefficient (WRC) analysis of the partial least squares model identified 11 important wavelengths. Simplified PLS-DA models were developed from two sets of reduced wavelengths selected by PCA and WRC and showed better performance with predictive correlation coefficients (R_p²) of 0.962 and 0.979, respectively, while PLS-DA with complete spectrum had R_p² of 0.943. Similarly, classification accuracy of PLS-DA was improved to 92.2% for WRC based predictive model. Calculation time was also reduced to 2.1 and 2.8 s for PCA and WRC based simplified PLS-DA models in comparison to the complete spectrum model that was taking 32.2 s on average to predict the classification of milk powder samples. These results demonstrated that HSI with appropriate data analysis methods could become a potential analyser for non-invasive testing of milk powder in the future.

Literature review: spectral imaging applied to poultry products

Consumption of poultry products is increasing worldwide, leading to an increased demand for safe, fresh, high-quality products. To ensure consumer safety and meet quality standards, poultry products must be routinely checked for fecal matter, food fraud, microbiological contamination, physical defects, and product quality. However, traditional screening methods are insufficient in providing real-time, nondestructive, chemical and spatial information about poultry products. Novel techniques, such as hyperspectral imaging (HSI), are being developed to acquire real-time chemical and spatial information about products without destruction of samples to ensure safety of products and prevent economic losses. This literature review provides a comprehensive overview of HSI applications to poultry products. The studies used for this review were found using the Google Scholar database by searching the following terms and their synonyms: “poultry” and “hyperspectral imaging”. A total of 67 studies were found to meet the criteria. After all relevant literature was compiled, studies were grouped into categories based on the specific material or characteristic of interest to be detected, identified, predicted, or quantified by HSI. Studies were found for each of the following categories: food fraud, fecal matter detection, microbiological contamination, physical defects, and product quality. Key findings and technological advancements were briefly summarized and presented for each category. Since the first application to poultry products 20 yr ago, HSI has been shown to be a successful alternative to traditional screening methods.

PTR-QiToF-MS and HSI for the characterization of fermented cocoa beans from different origins

The wide range of geographical cocoa production areas and the increasing consumption trend towards single origin products induced the necessity to verify and certify cocoa beans origin for quality assurance purposes. In this study cocoa beans of various origins were examined by machine olfaction and machine vision techniques. Fifty-nine fermented and dried Forastero cocoa beans from 23 different geographical origins (Africa, Americas, Southeast Asia) were investigated using Proton Transfer Reaction-Quadrupole interface-Time of Flight-Mass Spectrometry and Hyperspectral Imaging to elucidate the geographical information in the beans. The volatile and spectral fingerprints showed the same tendency in clustering samples from Africa separate from those from the Americas. High variability was observed for the Southeast Asian samples, which is most likely related to differences in fermentation. Machine olfaction and machine vision characterization provided a similar degree of separation but are complementary rapid techniques, which may be further developed for use in practical settings.

Developments and Challenges in Online NIR Spectroscopy for Meat Processing

Meat and meat products are popular foods due to their balanced nutritional nature and their availability in a variety of forms. In recent years, due to an increase in the consumer awareness regarding product quality and authenticity of food, rapid and effective quality control systems have been sought by meat industries. Near-Infrared (NIR) spectroscopy has been identified as a fast and cost-effective tool for estimating various meat quality parameters as well as detecting adulteration. This review focusses on the on/inline application of single and multiprobe NIR spectroscopy for the analysis of meat and meat products starting from the year 1996 to 2017. The article gives a brief description about the theory of NIR spectroscopy followed by its application for meat and meat products analysis. A detailed discussion is provided on the various studies regarding applications of NIR spectroscopy and specifically for on/inline monitoring along with their advantages and disadvantages. Additionally, a brief description has been given about the various chemometric techniques utilized in the mentioned studies. Finally, it discusses challenges encountered and future prospects of the technology. It is concluded that, advancements in the fields of NIR spectroscopy and chemometrics have immensely increased the potential of the technology as a reliable on/inline monitoring tool for the meat industry.

Identification of pesticide varieties by detecting characteristics of Chlorella pyrenoidosa using Visible/Near infrared hyperspectral imaging and Raman microspectroscopy technology

The main goal of this research is to examine the feasibility of applying Visible/Near-infrared hyperspectral imaging (Vis/NIR-HSI) and Raman microspectroscopy technology for non-destructive identification of pesticide varieties (glyphosate and butachlor). Both mentioned technologies were explored to investigate how internal elements or characteristics of Chlorella pyrenoidosa change when pesticides are applied, and in the meantime, to identify varieties of the pesticides during this procedure. Successive projections algorithm (SPA) was introduced to our study to identify seven most effective wavelengths. With those wavelengths suggested by SPA, a model of the linear discriminant analysis (LDA) was established to classify the pesticide varieties, and the correct classification rate of the SPA-LDA model reached as high as 100%. For the Raman technique, a few partial least squares discriminant analysis models were established with different preprocessing methods from which we also identified one processing approach that achieved the most optimal result. The sensitive wavelengths (SWs) which are related to algae's pigment were chosen, and a model of LDA was established with the correct identification reached a high level of 90.0%. The results showed that both Vis/NIR-HSI and Raman microspectroscopy techniques are capable to identify pesticide varieties in an indirect but effective way, and SPA is an effective wavelength extracting method. The SWs corresponding to microalgae pigments, which were influenced by pesticides, could also help to characterize different pesticide varieties and benefit the variety identification.

Monitoring of bacterial contamination on chicken meat surface using a novel narrowband spectral index derived from hyperspectral imagery data

This study presents a novel narrowband spectral index for monitoring bacterial contamination on chicken meat surface. Fresh chicken meats were prepared and stored aerobically in a refrigerator at 4°C for 11d. Hyperspectral images and the total viable count (TVC) of bacteria for meat samples were obtained every 24h. A new two band freshness index (TBFI) method was proposed for developing the bacteria prediction models. Results indicated that the model with the TBFI based on the wavelengths 650 and 700nm achieved the optimal estimation of TVC (R(2)=0.6833). The TBFI value for each image pixel was calculated using the above two wavelengths, and then used to predict the TVC for the corresponding pixel on the image. Finally, the predicted TVC were visualized to illustrate the temporal variation and spatial distribution of viable bacteria on meat surface over storage. The results demonstrate the promising potential of the developed TBFI for the detection of viable bacteria contamination on chicken meat surface.

Advances in feature selection methods for hyperspectral image processing in food industry applications: a review

There is an increased interest in the applications of hyperspectral imaging (HSI) for assessing food quality, safety, and authenticity. HSI provides abundance of spatial and spectral information from foods by combining both spectroscopy and imaging, resulting in hundreds of contiguous wavebands for each spatial position of food samples, also known as the curse of dimensionality. It is desirable to employ feature selection algorithms for decreasing computation burden and increasing predicting accuracy, which are especially relevant in the development of online applications. Recently, a variety of feature selection algorithms have been proposed that can be categorized into three groups based on the searching strategy namely complete search, heuristic search and random search. This review mainly introduced the fundamental of each algorithm, illustrated its applications in hyperspectral data analysis in the food field, and discussed the advantages and disadvantages of these algorithms. It is hoped that this review should provide a guideline for feature selections and data processing in the future development of hyperspectral imaging technique in foods.

Detection of early blight and late blight diseases on tomato leaves using hyperspectral imaging

This study investigated the potential of using hyperspectral imaging for detecting different diseases on tomato leaves. One hundred and twenty healthy, one hundred and twenty early blight and seventy late blight diseased leaves were selected to obtain hyperspectral images covering spectral wavelengths from 380 to 1023 nm. An extreme learning machine (ELM) classifier model was established based on full wavelengths. Successive projections algorithm (SPA) was used to identify the most important wavelengths. Based on the five selected wavelengths (442, 508, 573, 696 and 715 nm), an ELM model was re-established. Then, eight texture features (mean, variance, homogeneity, contrast, dissimilarity, entropy, second moment and correlation) based on gray level co-occurrence matrix (GLCM) at the five effective wavelengths were extracted to establish detection models. Among the models which were established based on spectral information, all performed excellently with the overall classification accuracy ranging from 97.1% to 100% in testing sets. Among the eight texture features, dissimilarity, second moment and entropy carried most of the effective information with the classification accuracy of 71.8%, 70.9% and 69.9% in the ELM models. The results demonstrated that hyperspectral imaging has the potential as a non-invasive method to identify early blight and late blight diseases on tomato leaves.

Identification of species and geographical strains of Sitophilus oryzae and Sitophilus zeamais using the visible/near-infrared hyperspectral imaging technique

BACKGROUND

Identifying stored-product insects is essential for granary management. Automated, computer-based classification methods are rapidly developing in many areas. A hyperspectral imaging technique could potentially be developed to identify stored-product insect species and geographical strains. This study tested and adapted the technique using four geographical strains of each of two insect species, the rice weevil and maize weevil, to collect and analyse the resultant hyperspectral data.

RESULTS

Three characteristic images that corresponded to the dominant wavelengths, 505, 659 and 955 nm, were selected by multivariate image analysis. Each image was processed, and 22 morphological and textural features from regions of interest were extracted as the inputs for an identification model. We found the backpropagation neural network model to be the superior method for distinguishing between the insect species and geographical strains. The overall recognition rates of the classification model for insect species were 100 and 98.13% for the calibration and prediction sets respectively, while the rates of the model for geographical strains were 94.17 and 86.88% respectively.

CONCLUSION

This study has demonstrated that hyperspectral imaging, together with the appropriate recognition method, could provide a potential instrument for identifying insects and could become a useful tool for identification of Sitophilus oryzae and Sitophilus zeamais to aid in the management of stored-product insects.