Search

Ensuring the quality of medicinal herbs in the herbal market is crucial. However, the genetic and physical similarities among medicinal materials have led to issues of mixing and counterfeit distribution, posing significant challenges to quality assurance. Recent advancements in deep learning technology, widely applied in the field of computer vision, have demonstrated the potential to classify images quickly and accurately, even those that can only be distinguished by experts. This study aimed to develop a classification model based on deep learning technology to distinguish RGB images of seeds from Ziziphus jujuba Mill. var. spinosa, Ziziphus mauritiana Lam., and Hovenia dulcis Thunb. Using three advanced convolutional neural network (CNN) architectures—ResNet-50, Inception-v3, and DenseNet-121—all models demonstrated a classification performance above 98% on the test set, with classification times as low as 23 ms. These results validate that the models and methods developed in this study can effectively distinguish Z. jujuba seeds from morphologically similar species. Furthermore, the strong performance and speed of these models make them suitable for practical use in quality inspection settings.

ABSTRACT Latilactobacillus sakei is an important bacterial species used as a starter culture for fermented foods; however, two subspecies within this species exhibit different properties in the foods. Matrix-assisted laser desorption/ionization-time of flight mass spectrometer (MALDI-TOF MS) is the gold standard for microbial fingerprinting. However, the resolution power is down to the species level. This study was to combine MALDI-TOF mass spectra and machine learning to develop a new method to identify two L. sakei subspecies (L. sakei subsp. sakei and L. sakei subsp. carnosus) and non-L. sakei species. Totally, 227 strains were collected, with 908 spectra obtained via on- and off-plate protein extraction. Only 68.7% of strains were correctly identified at the subspecies level in the Biotyper database; however, a high level of performance was observed from the machine learning models. Partial least squares-discriminant analysis (PLS-DA), principal component analysis-K-nearest neighbor (PCA-KNN), and support vector machine (SVM) demonstrated 0.823, 0.914, and 0.903 accuracies, respectively, whereas the random forest (RF) achieved an accuracy of 0.954, with an area under the receiver operating characteristic (AUROC) curve of 0.99, outperforming the other algorithms in distinguishing the subspecies. The machine learning proved to be a promising technique for the rapid and high-resolution classification of L. sakei subspecies using MALDI-TOF MS.IMPORTANCELatilactobacillus sakei plays a significant role in the realm of food bacteria. One particular subspecies of L. sakei is employed as a protective agent during food fermentation, whereas another strain is responsible for food spoilage. Hence, it is crucial to precisely differentiate between the two subspecies of L. sakei. In this study, machine learning models based on protein mass peaks were developed for the first time to distinguish L. sakei subspecies. Furthermore, the efficacy of three commonly used machine learning algorithms for microbial classification was evaluated. Our results provide the foundation for future research on developing machine learning models for the classification of microbial species or subspecies. In addition, the developed model can be used in the food industry to monitor L. sakei subspecies in fermented foods in a time- and cost-effective method for food quality and safety.

The rapid and accurate classification of aquatic products is crucial for ensuring food safety, production efficiency, and economic benefits. However, traditional manual methods for classifying ark shell species based on phenotype are time-consuming and inefficient, especially during peak seasons when the demand is high and labor is scarce. This study aimed to develop a deep learning model for the automated identification and classification of commercially important three ark shells (Tegillarca granosa, Anadara broughtonii, and Anadara kagoshimensis) from images. The ark shells were collected and identified using a polymerase chain reaction method developed in a previous study, and a total of 1,400 images were categorized into three species. Three convolutional neural network (CNN) models, Visual Geometry Group Network (VGGnet), Inception-Residual Network (ResNet), and SqueezeNet, were then applied to two different classification sets, Set-1 (four bivalve species) and Set-2 (three ark shell species). Our results showed that SqueezeNet demonstrated the highest accuracy during the training phase for both classification sets, whereas Inception-ResNet exhibited superior accuracy during the validation phase. Similar results were obtained after developing a third classification set (Set-3) to classify six categories by combining Set-1 and Set-2. Overall, the developed CNN-based classification model exhibited a performance comparable or superior to that presented in previous studies and can provide a theoretical basis for bivalve classification, thereby contributing to improved food safety, production efficiency, and economic benefits in the aquatic products industry.

The maintenance of ion balance in closed hydroponic solutions is essential to improve the crop quality and recycling efficiency of nutrient solutions. However, the absence of robust ion sensors for key ions such as P and Mg and the coupling of ions in fertilizer salts render it difficult to effectively manage ion-specific nutrient solutions. Although ion-specific dosing algorithms have been established, their effectiveness has been inadequately explored. In this study, a decision-tree-based dosing algorithm was developed to calculate the optimal volumes of individual nutrient stock solutions to be supplied for five major nutrient ions, i.e., NO3, K, Ca, P, and Mg, based on the concentrations of NO3, K, and Ca and remaining volume of the recycled nutrient solution. In the performance assessment based on five nutrient solution samples encompassing the typical concentration ranges for leafy vegetable cultivation, the ion-selective electrode array demonstrated feasible accuracies, with root mean square errors of 29.5, 10.1, and 6.1 mg·L-1 for NO3, K, and Ca, respectively. In a five-step replenishment test involving varying target concentrations and nutrient solution volumes, the system formulated nutrient solutions according to the specified targets, exhibiting average relative errors of 10.6 ± 8.0%, 7.9 ± 2.1%, 8.0 ± 11.0%, and 4.2 ± 3.7% for the Ca, K, and NO3 concentrations and volume of the nutrient solution, respectively. Furthermore, the decision tree method helped reduce the total fertilizer injections and carbon emissions by 12.8% and 20.6% in the stepwise test, respectively. The findings demonstrate that the decision-tree-based dosing algorithm not only enables more efficient reuse of nutrient solution compared to the existing simplex method but also confirms the potential for reducing carbon emissions, indicating the possibility of sustainable agricultural development.

The greenhouse environment plays a crucial role in providing favorable conditions for crop growth, significantly improving their quality and yield. Accurate prediction of greenhouse environmental factors is essential for their effective control. Although artificial intelligence technologies for predicting greenhouse environments have been researched recently, there are limitations in applying these to general greenhouse environments due to computing resources or issues with interpretability. Moreover, research on environmental prediction models specifically for melon greenhouses is also lacking. In this study, machine learning models based on MLR (Multiple Linear Regression), SVM (Support Vector Machine), ANN (Artificial Neural Network), and XGBoost were developed to predict the internal temperature, relative humidity, and CO2 conditions of melon greenhouses 30 min in advance. The XGBoost model demonstrated high accuracy and stability, with an R2 value of up to 0.9929 and an RPD (Residual Predictive Deviation) of 11.8464. Furthermore, the analysis of the XGBoost model’s feature importance and decision trees revealed that the model learned the complex relationships and impacts among greenhouse environmental factors. In conclusion, this study successfully developed a predictive model for a greenhouse environment for melon cultivation. The model developed in this study can facilitate an understanding and efficient management of the greenhouse environment, contributing to improvements in crop yield and quality.

Cynanchum wilfordii is a perennial tuberous root in the Asclepiadaceae family that has long been used medicinally. Although C. wilfordii is distinct in origin and content from Cynancum auriculatum, a genus of the same species, it is difficult for the public to recognize because the ripe fruit and root are remarkably similar. In this study, images were collected to categorize C. wilfordii and C. auriculatum, which were then processed and input into a deep-learning classification model to corroborate the results. By obtaining 200 photographs of each of the two cross sections of each medicinal material, approximately 800 images were employed, and approximately 3200 images were used to construct a deep-learning classification model via image augmentation. For the classification, the structures of Inception-ResNet and VGGnet-19 among convolutional neural network (CNN) models were used, with Inception-ResNet outperforming VGGnet-19 in terms of performance and learning speed. The validation set confirmed a strong classification performance of approximately 0.862. Furthermore, explanatory properties were added to the deep-learning model using local interpretable model-agnostic explanation (LIME), and the suitability of the LIME domain was assessed using cross-validation in both situations. Thus, artificial intelligence may be used as an auxiliary metric in the sensory evaluation of medicinal materials in future, owing to its explanatory ability.

Abstract Background On tomato plants, the flowering truss is a group or cluster of smaller stems where flowers and fruit develop, while the growing truss is the most extended part of the stem. Because the state of the growing truss reacts sensitively to the surrounding environment, it is essential to control its growth in the early stages. With the recent development of information and artificial intelligence technology in agriculture, a previous study developed a real-time acquisition and evaluation method for images using robots. Furthermore, we used image processing to locate the growing truss to extract growth information. Among the different vision algorithms, the CycleGAN algorithm was used to generate and transform unpaired images using generated learning images. In this study, we developed a robot-based system for simultaneously acquiring RGB and depth images of the growing truss of the tomato plant. Results The segmentation performance for approximately 35 samples was compared via false negative (FN) and false positive (FP) indicators. For the depth camera image, we obtained FN and FP values of 17.55 ± 3.01% and 17.76 ± 3.55%, respectively. For the CycleGAN algorithm, we obtained FN and FP values of 19.24 ± 1.45% and 18.24 ± 1.54%, respectively. When segmentation was performed via image processing through depth image and CycleGAN, the mean intersection over union (mIoU) was 63.56 ± 8.44% and 69.25 ± 4.42%, respectively, indicating that the CycleGAN algorithm can identify the desired growing truss of the tomato plant with high precision. Conclusions The on-site possibility of the image extraction technique using CycleGAN was confirmed when the image scanning robot drove in a straight line through a tomato greenhouse. In the future, the proposed approach is expected to be used in vision technology to scan tomato growth indicators in greenhouses using an unmanned robot platform.

The integration of hyperspectral imaging with machine learning algorithms has presented a promising strategy for the non-invasive and rapid detection of plant metabolites. For this study, we developed prediction models using partial least squares regression (PLSR) and boosting algo-rithms (such as AdaBoost, XGBoost, and LightGBM) for five metabolites in Brassica juncea leaves: total chlorophyll, phenolics, flavonoids, glucosinolates, and anthocyanins. To enhance the model performance, we employed several spectral data preprocessing methods and feature-selection al-gorithms. Our results showed that the boosting algorithms generally outperformed the PLSR models in terms of prediction accuracy. In particular, the LightGBM model for chlorophyll and the AdaBoost model for flavonoids improved the prediction performance, with R2p = 0.71–0.74, com-pared to the PLSR models (R2p = 0.53–0.58). The final models for the glucosinolates and anthocya-nins performed sufficiently for practical uses such as screening, with R2p = 0.82–0.85 and RPD = 2.4–2.6. Our findings indicate that the application of a single preprocessing method is more effective than utilizing multiple techniques. Additionally, the boosting algorithms with feature selection ex-hibited superior performance compared to the PLSR models in the majority of cases. These results highlight the potential of hyperspectral imaging and machine learning algorithms for the non-destructive and rapid detection of plant metabolites, which could have significant implications for the field of smart agriculture.

Gray mold disease is one of the most frequently occurring diseases in strawberries. Given that it spreads rapidly, rapid countermeasures are necessary through the development of early diagnosis technology. In this study, hyperspectral images of strawberry leaves that were inoculated with gray mold fungus to cause disease were taken; these images were classified into healthy and infected areas as seen by the naked eye. The areas where the infection spread after time elapsed were classified as the asymptomatic class. Square regions of interest (ROIs) with a dimensionality of 16 × 16 × 150 were acquired as training data, including infected, asymptomatic, and healthy areas. Then, 2D and 3D data were used in the development of a convolutional neural network (CNN) classification model. An effective wavelength analysis was performed before the development of the CNN model. Further, the classification model that was developed with 2D training data showed a classification accuracy of 0.74, while the model that used 3D data acquired an accuracy of 0.84; this indicated that the 3D data produced slightly better performance. When performing classification between healthy and asymptomatic areas for developing early diagnosis technology, the two CNN models showed a classification accuracy of 0.73 with regards to the asymptomatic ones. To increase accuracy in classifying asymptomatic areas, a model was developed by smoothing the spectrum data and expanding the first and second derivatives; the results showed that it was possible to increase the asymptomatic classification accuracy to 0.77 and reduce the misclassification of asymptomatic areas as healthy areas. Based on these results, it is concluded that the proposed 3D CNN classification model can be used as an early diagnosis sensor of gray mold diseases since it produces immediate on-site analysis results of hyperspectral images of leaves.

The greenhouse industry achieves stable agricultural production worldwide. Various information and communication technology techniques to model and control the environment have been applied as data from environmental sensors and actuators in greenhouses are monitored in real time. The current study designed data-based, deep learning models for evapotranspiration (ET) and humidity in tomato greenhouses. Using time-series data and applying long short-term memory (LSTM) modeling, an ET prediction model was developed and validated in comparison with the Stanghellini model. Training with 20-day and testing with 3-day data resulted in RMSEs of 0.00317 and 0.00356 kgm−2 s−1, respectively. The standard error of prediction indicated errors of 5.76 and 6.45% in training and testing, respectively. Variables were used to produce a feature map using a two-dimensional convolution layer which was transferred to a subsequent layer and finally connected with the LSTM structure for modeling. The RMSE in humidity prediction using the test dataset was 2.87, indicating a performance better than conventional RNN-LSTM models. Irrigation plans and humidity control may be more accurately conducted in greenhouse cultivation using this model.

Partial least squares regression (PLSR) prediction models were developed using hyperspectral imaging for noninvasive detection of the five most representative functional components in Brassica juncea leaves: chlorophyll, carotenoid, phenolic, glucosinolate, and anthocyanin contents. The region of interest for functional component analysis was chosen by polygon selection and the extracted average spectra were used for model development. For pre-processing, 10 combinations of Savitzky–Golay filter (S. G. filter), standard normal variate (SNV), multiplicative scatter correction (MSC), 1st-order derivative (1st-Der), 2nd-order derivative (2nd-Der), and normalization were applied. Root mean square errors of calibration (RMSEP) was used to assess the performance accuracy of the constructed prediction models. The prediction model for total anthocyanins exhibited the highest prediction level (RV2 = 0.8273; RMSEP = 2.4277). Pre-processing combination of SNV and 1st-Der with spectral data resulted in high-performance prediction models for total chlorophyll, carotenoid, and glucosinolate contents. Pre-processing combination of S. G. filter and SNV gave the highest prediction rate for total phenolics. SNV inclusion in the pre-processing conditions was essential for developing high-performance accurate prediction models for functional components. By enabling visualization of the distribution of functional components on the hyperspectral images, PLSR prediction models will prove valuable in determining the harvest time.

The detection of nitrate pollutants is a widely used strategy for protecting water sources. Although ion-selective electrodes (ISEs) have been considered for the determination of ion concentrations in water, the accuracy of ISE technology decreases owing to the signal drift and decreasing sensitivity over time. The objectives of the present study were: (1) to develop an online water monitoring system mainly consisting of an Arduino board-based Internet-of-Things (IoT) device and nitrate ISEs; and (2) to propose a self-diagnostic function for monitoring and reporting the condition of the ISEs. The developed system communicates with the cloud server by using the message queuing telemetry transport (MQTT) protocol and provides monitoring information through the developed cloud-based webpage. In addition, the online monitoring system provides information on the electrode status, which is determined based on a self-diagnostic index (SDI, with a range of 0–100) of the electrode drift and sensitivity. The diagnostic method for monitoring and reporting the electrode status was validated in a one-month-long laboratory test followed by a field test in a stream near an agricultural facility. Moreover, a self-diagnostic index (SDI) was applied in the final field experiments with an accuracy of 0.77.

The priority placed on animal welfare in the meat industry is increasing the importance of understanding livestock behavior. In this study, we developed a web-based monitoring and recording system based on artificial intelligence analysis for the classification of cattle sounds. The deep learning classification model of the system is a convolutional neural network (CNN) model that takes voice information converted to Mel-frequency cepstral coefficients (MFCCs) as input. The CNN model first achieved an accuracy of 91.38% in recognizing cattle sounds. Further, short-time Fourier transform-based noise filtering was applied to remove background noise, improving the classification model accuracy to 94.18%. Categorized cattle voices were then classified into four classes, and a total of 897 classification records were acquired for the classification model development. A final accuracy of 81.96% was obtained for the model. Our proposed web-based platform that provides information obtained from a total of 12 sound sensors provides cattle vocalization monitoring in real time, enabling farm owners to determine the status of their cattle.

Maintaining environmental conditions for proper plant growth in greenhouses requires managing a variety of factors; ventilation is particularly important because inside temperatures can rise rapidly in warm climates. The structure of the window installed in a greenhouse is very diverse, and it is difficult to identify the characteristics that affect the temperature inside the greenhouse when multiple windows are driven, respectively. In this study, a new ventilation control logic using an output feedback neural-network (OFNN) prediction and optimization method was developed, and this approach was tested in multi-window greenhouses used for strawberry production. The developed prediction model used 15 inputs and achieved a highly accurate performance (R2 of 0.94). In addition, the method using an algorithm based on an OFNN was proposed for optimizing considered six window-opening behavior. Three case studies confirmed the optimization performance of OFNN in the nonlinear model and verified the performance through simulations. Finally, a control system based on this logic was used in a field experiment for six days by comparing two greenhouses driven by conventional control logic and the developed control logic; a comparison of the results showed RMSEs of 3.01 °C and 2.45 °C, respectively. It confirmed the improved control performance in comparison to a conventional ventilation control system.

Phosphate is a key element affecting plant growth. Therefore, the accurate determination of phosphate concentration in hydroponic nutrient solutions is essential for providing a balanced set of nutrients to plants within a suitable range. This study aimed to develop a data fusion approach for determining phosphate concentrations in a paprika nutrient solution. As a conventional multivariate analysis approach using spectral data, partial least squares regression (PLSR) and principal components regression (PCR) models were developed using 56 samples for calibration and 24 samples for evaluation. The R2 values of estimation models using PCR and PLSR ranged from 0.44 to 0.64. Furthermore, an estimation model using raw electromotive force (EMF) data from cobalt electrodes gave R2 values of 0.58–0.71. To improve the model performance, a data fusion method was developed to estimate phosphate concentration using near infrared (NIR) spectral and cobalt electrochemical data. Raw EMF data from cobalt electrodes and principle component values from the spectral data were combined. Results of calibration and evaluation tests using an artificial neural network estimation model showed that R2 = 0.90 and 0.89 and root mean square error (RMSE) = 96.70 and 119.50 mg/L, respectively. These values are sufficiently high for application to measuring phosphate concentration in hydroponic solutions.