Search

The variety and content of high-quality proteins in sunflower seeds are higher than those in other cereals. However, sunflower seeds can suffer from abnormalities, such as breakage and deformity, during planting and harvesting, which hinder the development of the sunflower seed industry. Traditional methods such as manual sensory and machine sorting are highly subjective and cannot detect the internal characteristics of sunflower seeds. The development of spectral imaging technology has facilitated the application of terahertz waves in the quality inspection of sunflower seeds, owing to its advantages of non-destructive penetration and fast imaging. This paper proposes a novel terahertz image classification model, MobileViT-E, which is trained and validated on a self-constructed dataset of sunflower seeds. The results show that the overall recognition accuracy of the proposed model can reach 96.30%, which is 4.85%, 3%, 7.84% and 1.86% higher than those of the ResNet-50, EfficientNeT, MobileOne and MobileViT models, respectively. At the same time, the performance indices such as the recognition accuracy, the recall and the F1-score values are also effectively improved. Therefore, the MobileViT-E model proposed in this study can improve the classification and identification of normal, damaged and deformed sunflower seeds, and provide technical support for the non-destructive detection of sunflower seed quality.

The detection of concealed suspicious objects in public places is a critical issue and a popular research topic. Terahertz (THz) imaging technology, as an emerging detection method, can penetrate materials without emitting ionizing radiation, providing a new approach to detecting concealed suspicious objects. This study focuses on the detection of concealed suspicious objects wrapped in different materials such as polyethylene and kraft paper, including items like scissors, pistols, and blades, using THz imaging technology. To address issues such as the lack of texture details in THz images and the contour similarity of different objects, which can lead to missed detections and false alarms, we propose a THz concealed suspicious object detection model based on SMR–YOLO (SPD_Mobile + RFB + YOLO). This model, based on the MobileNext network, introduces the spatial-to-depth convolution (SPD-Conv) module to replace the backbone network, reducing computational and parameter load. The inclusion of the receptive field block (RFB) module, which uses a multi-branch structure of dilated convolutions, enhances the network’s depth features. Using the EIOU loss function to assess the accuracy of predicted box localization further optimizes convergence speed and localization accuracy. Experimental results show that the improved model achieved mAP@0.5 and mAP@0.5:0.95 scores of 98.9% and 89.4%, respectively, representing improvements of 0.2% and 1.8% over the baseline model. Additionally, the detection speed reached 108.7 FPS, an improvement of 23.2 FPS over the baseline model. The model effectively identifies concealed suspicious objects within packages, offering a novel approach for detection in public places.

A terahertz metamaterial microfluidic sensing chip for ultrasensitive detection is proposed to investigate the response of substances to terahertz radiation in liquid environments and enhance the molecular fingerprinting of trace substances. The structure consists of a cover layer, a metal microstructure, a microfluidic channel, a metal reflective layer, and a buffer layer from top to bottom, respectively. The simulation results show that there are three obvious resonance absorption peaks in the range of 1.5–3.0 THz and the absorption intensities are all above 90%. Among them, the absorption intensity at M1 = 1.971 THz is 99.99%, which is close to the perfect absorption, and its refractive index sensitivity and Q-factor are 859 GHz/RIU and 23, respectively, showing excellent sensing characteristics. In addition, impedance matching and equivalent circuit theory are introduced in this paper to further analyze the physical mechanism of the sensor. Finally, we perform numerical simulations using refractive index data of normal and cancer cells, and the results show that the sensor can distinguish different types of cells well. The chip can reduce the sample pretreatment time as well as enhance the interaction between terahertz waves and matter, which can be used for early disease screening and food quality and safety detection in the future.

In order to achieve accurate identification of rice varieties, a rice variety identification method based on terahertz time-domain spectroscopy (THz-TDS) technology was proposed. To select the terahertz band by Interval partial least square (iPLS) and standard deviation (SD), which determined the absorption spectrum data of the 0.53~1.21 THz band as the input data of the classification model. Then to use the decision tree (DT) identified absorption spectra of four kinds of rice and the model parameters was obtained by combining with the grid search algorithm. In order to increase the experimental contrast, logistic regression models and support vector machine models were used separately for comparative experiments, and the model classification accuracy was 80.75% and 88.75%, respectively. Experimental results showed that the terahertz time-domain spectroscopy technology combined with SD, iPLS and DT methods can realize the accurate identification of rice varieties with an accuracy rate of up to 95%, providing a new identification method for the identification of agricultural varieties.

With the strengthening of worldwide counter-terrorism initiatives, it is increasingly important to detect contrabands such as controlled knives and flammable materials hidden in clothes and bags. Terahertz (THz) imaging technology is widely used in the field of contraband detection due to its advantages of high imaging speed and strong penetration. However, the terahertz images are of poor qualities and lack texture details. Traditional target detection methods suffer from low detection speeds, misdetection, and omission of contraband. This work pre-processes the original dataset using a variety of image processing methods and validates the effect of these methods on the detection results of YOLOv7. Meanwhile, the lightweight and multi-object detection YOLOv7 (LWMD-YOLOv7) algorithm is proposed. Firstly, to meet the demand of real-time for multi-target detection, we propose the space-to-depth mobile (SPD_Mobile) network as the lightweight feature extraction network. Secondly, the selective attention module large selective kernel (LSK) network is integrated into the output of the multi-scale feature map of the LWMD-YOLOv7 network, which enhances the effect of feature fusion and strengthens the network’s attention to salient features. Finally, Distance Intersection over Union (DIOU) is used as the loss function to accelerate the convergence of the model and to have a better localisation effect for small targets. The experimental results show that the YOLOv7 algorithm achieves the best detection results on the terahertz image dataset after the non-local mean filtering process. The LWMD-YOLOv7 algorithm achieves a detection accuracy P of 98.5%, a recall R of 97.5%, and a detection speed of 112.4 FPS, which is 26.9 FPS higher than that of the YOLOv7 base network. The LWMD-YOLOv7 achieves a better balance between detection accuracy and detection speed. It provides a technological reference for the automated detection of contraband in terahertz images.

Abstract The Tunpu settlements are the historical product of the co-evolution development of the unique geographical and cultural environment of Guizhou and the military defense policies of Ming Dynasty (1368–1644 AD), which created products such as the station troops and station farms. In this study, taking 10 Tunpu settlements in central Guizhou as the research objects, the space diversity of the Tunpu settlement and its influencing factors were analyzed based on the theory and calculation method of plant community species diversity. The results showed that: (1) On the whole, the space α-diversities of the military Tunpu settlements and the commercial Tunpu settlements were relatively high, with strong spatial similarity, while the residential Tunpu settlements had a lower degree of space α-diversity and significant difference. The space α-diversity of different types of the Tunpu settlement presents a variety of changes from the core protected areas to coordinated control areas. (2) There was a significant correlation between the spatial network and spatial morphological indicators and space β-diversity among the Tunpu settlement, and the stepwise regression showed that spatial morphological indicators had a strong explanatory power for the space β-diversity among the Tunpu settlements. (3) The space diversities of the Tunpu settlements were affected by the multiple synergies of natural environment, policy support, social-economic factors and cultural customs. Among them, policy factors and cultural factors were the dominant factors in the space diversity characteristics of the Tunpu settlements. By analyzing the space diversity law and its driving factors of the Tunpu settlements, this study could provide an important scientific basis for the protection planning and management of traditional settlements.

Cereal and oil video surveillance data play a vital role in food traceability, which not only helps to ensure the quality and safety of food, but also helps to improve the efficiency and transparency of the supply chain. Traditional video surveillance systems mainly adopt a centralized storage mode, which is characterized by the deployment of multiple monitoring nodes and a large amount of data storage. It is difficult to guarantee the data security, and there is an urgent need for a solution that can achieve the safe and efficient storage of cereal and oil video surveillance data. This study proposes a blockchain-based abnormal data storage model for cereal and oil video surveillance. The model introduces a deep learning algorithm to process the cereal and oil video surveillance data, obtaining images with abnormal behavior from the monitoring data. The data are stored on a blockchain after hash operation, and InterPlanetary File System (IPFS) is used as a secondary database to store video data and alleviate the storage pressure on the blockchain. The experimental results show that the model achieves the safe and efficient storage of cereal and oil video surveillance data, providing strong support for the sustainable development of the cereal and oil industry.

Terahertz (THz) radiation ( $0.1\sim 10$ THz) shows great potential in agricultural products detection, biomedical, and security inspection in recent years. Machine learning methods are widely used to support the user demand of higher efficiency and high prediction accuracy. The technological and key challenges of machine learning methods are for THz spectroscopy and image data preprocessing, reconstruction algorithms, and qualitative and quantitative analysis. In this paper, an exhaustive review of recent related works of THz detection and imaging techniques and machine learning methods are presented. The application of machine learning methods combined with THz technology in quality inspection of agricultural products, biomedical, security inspection, and materials science are highlighted. Challenges of machine learning methods for these applications are addressed. The development trend and future perspectives of THz technology are also discussed.

The structure of the grain-and-oil-food-supply chain has the characteristics of complexity, cross-regionality, a long cycle, and numerous participants, making it difficult to maintain the safety of supply. In recent years, some phenomena have emerged in the field of grain procurement and sale, such as topping the new with the old, rotating grains, the pressure of grades and prices, and counterfeit oil food, which have seriously threatened grain-and-oil-food security. Blockchain technology has the advantage of decentralization and non-tampering Therefore, this study analyzes the characteristics of traceability data in the grain-and-oil-food-supply chain, and presents a blockchain-based traceability model for the grain-and-oil-food-supply chain. Firstly, a new method combining blockchain and machine learning is proposed to enhance the authenticity and reliability of blockchain source data by constructing anomalous data-processing models. In addition, a lightweight blockchain-storage method and a data-recovery mechanism are proposed to reduce the pressure on supply-chain-data storage and improve fault tolerance. The results indicate that the average query delay of public data is 0.42 s, the average query delay of private data is 0.88 s, and the average data-recovery delay is 1.2 s. Finally, a blockchain-based grain-and-oil-food-supply-chain traceability system is designed and built using Hyperledger Fabric. Compared with the existing grain-and-oil-food-supply chain, the model constructed achieves multi-source heterogeneous data uploading, lightweight storage, data recovery, and traceability in the supply chain, which are of great significance for ensuring the safety of grain-and-oil food in China.

In order to solve the problem of large image noise and unremarkable features caused by factors such as fluctuations in the power of a light source during the terahertz image acquisition of wheat, this paper proposes a THz image-denoising model called G-RRDB. Firstly, a module called Ghost-LKA is proposed by combining a large kernel convolutional attention mechanism module with a Ghost convolutional structure, which improves the characteristics of the network to acquire a global sensory field. Secondly, by integrating a spatial attention mechanism with channel attention, an attention module called DAB is proposed to enhance the network’s attention to important features. Thirdly, the Ghost-LKA module and DAB module are combined with the baseline model, thus proposing the dense residual denoising network G-RRDB. Compared with traditional denoising networks, both the PSNR and SSIM are improved. The prediction accuracy of G-RRDB is verified through the classification of the VGG16 network, achieving a rate of 92.8%, which represents an improvement of 1.7% and 0.2% compared to the denoised images obtained from the baseline model and the combined baseline model with the DAB module, respectively. The experimental results demonstrate that G-RRDB, a THz image-denoising model based on dense residual structure for moldy wheat, exhibits excellent denoising performance.

A terahertz metamaterial comprised of an array of cross rectangular split-ring resonators (CRSRR) was proposed and analyzed for sensing applications, and it exhibited two resonances in the frequency range of 0.2–3 THz. The resonant frequencies of different resonant modes were explained using equivalent circuit models. Furthermore, the influence on equivalent capacitance and inductance of the circuit with respect to different geometrical dimensions of the CRSRR structure were analyzed, and the results indicated that the resonant frequencies of the proposed metamaterial can be designed as the desired value by adjusting the CRSRR unit geometry. In addition, the sensing performances of the metamaterial were calculated based on the optimized structure, showing that it had high refractive index sensitivity of 309 and 730 GHz/RIU at two resonant frequencies, respectively. Meanwhile, such ability to operate at two frequency bands enabled the designed sensor could characterize the identical samples at different frequencies, thereby increasing the sensing sensitivity and decreasing the impact of environmental disturbance. Our study opens up new prospects in the design of terahertz metamaterial sensors with high sensitivity in a multi-band range, which is essential to meet increasing needs in terahertz sensing.

Retention of rice freshness is highly dependent on storage temperature. Timely and accurate identification of rice freshness is critical to ensure food security. Here, we characterize the freshness of rice in reference to storage temperature. Terahertz reflection imaging is a non-destructive and deeply penetrating technique that can be used for detecting rice freshness. Due to the shortcomings of traditional machine learning, such as limited processing of nonlinear problems and insufficient computing power. Deep learning has the advantages of strong learning ability and high portability. Therefore, for rice freshness identification, the VGG19 network and the Inception-ResNet-v2 network were used in this paper. Moreover, we propose an improved 1D-VGG19-Inception-ResNet-A network. This network possesses the advantages of low time consumption from the 1D-VGG19 network and high classification accuracy from the 1D-Inception-ResNet-V2 network. Compared with the traditional algorithms, the accuracy of the proposed network is significantly improved, with the rice freshness recognition accuracy of 99.80%. The experimental results indicate that terahertz spectral imaging and deep learning algorithms are viable tools for monitoring rice freshness.

To solve the problems existing in traditional biochemical methods, such as complex sample pretreatment requirements, tedious detection processes and low detection accuracies with respect to rice species and adulteration, the volatile flavor substances of five kinds of rice are detected using headspace-gas chromatography-ion mobility spectrometry (HGC-IMS) to effectively identify the quality of rice and adulterated rice. The ion migration fingerprint spectra of five kinds of rice are identified using a semi-supervised generative adversarial network (SSGAN). We replace the output layer of the discriminator in a GAN with a softmax classifier, thus extending the GAN to a semi-supervised GAN. We define additional category tags for generated samples to guide the training process. Semi-supervised training is used to optimize the network parameters, and the trained discriminant network is used for classifying HGC-IMS images. The experimental results show that the prediction accuracy of the model reaches 98.00%, which is significantly higher than the rates achieved by other models, such as a decision tree, a support vector machine (SVM), improved SVM models (LS-SVM and PCA-SVM) and local geometric structure Fisher analysis (LGSFA); 98.00% is also higher than the prediction accuracies of the VGGNet, ResNet and Fast RCNN deep learning models. The experimental results also show that the accuracy of HGC-IMS image classification for identifying adulterated rice reaches 97.30%, which is higher than those of traditional chromatographic or spectral methods. The proposed method overcomes the shortcomings of some intelligent algorithms regarding the application of ion migration spectra and is feasible for accurately predicting rice varieties and adulterated rice.

The absence of labeled samples limits the development of speech emotion recognition (SER). Data augmentation is an effective way to address sample sparsity. However, there is a lack of research on data augmentation algorithms in the field of SER. In this paper, the effectiveness of classical acoustic data augmentation methods in SER is analyzed, based on which a strong generalized speech emotion recognition model based on effective data augmentation is proposed. The model uses a multi-channel feature extractor consisting of multiple sub-networks to extract emotional representations. Different kinds of augmented data that can effectively improve SER performance are fed into the sub-networks, and the emotional representations are obtained by the weighted fusion of the output feature maps of each sub-network. And in order to make the model robust to unseen speakers, we employ adversarial training to generalize emotion representations. A discriminator is used to estimate the Wasserstein distance between the feature distributions of different speakers and to force the feature extractor to learn the speaker-invariant emotional representations by adversarial training. The simulation experimental results on the IEMOCAP corpus show that the performance of the proposed method is 2–9% ahead of the related SER algorithm, which proves the effectiveness of the proposed method.

Terahertz (THz)-detection technology has been proven to be an effective and rapid non-destructive detection approach in biomedicine, quality control, and safety inspection, among other applications. However, the sensitivity of such a detection method is limited due to the insufficient power of the terahertz source and the low content, or ambiguous characteristics, of the analytes to be measured. Metamaterial (MM) is an artificial structure in which periodic sub-wavelength units are arranged in a regular manner, resulting in extraordinary characteristics beyond those possessed by natural materials. It is an effective method to improve the ability of terahertz spectroscopy detection by utilizing the metamaterial as a sensor. In this paper, a dual-band, high-sensitivity THz MM sensor based on the split metal stacking ring resonator (SMSRR) is proposed. The appliance exhibited two resonances at 0.97 and 2.88 THz in the range of 0.1 to 3 THz, realizing multi-point matching between the resonance frequency and the characteristic frequency of the analytes, which was able to improve the reliability and detection sensitivity of the system. The proposed sensor has good sensing performance at both resonant frequencies and can achieve highest sensitivities of 304 GHz/RIU and 912 GHz/RIU with an appropriate thickness of the analyte. Meanwhile, the advantage of multi-point matching of the proposed sensor has been validated by distinguishing four edible oils based on their different refractive indices and demonstrating that the characteristics obtained in different resonant frequency bands are consistent. This work serves as a foundation for future research on band extension and multi-point feature matching in terahertz detection, potentially paving the way for the development of high-sensitivity THz MM sensors.

This paper offers a prospective solution to the poor quality and less prominent features of the original terahertz spectral images of unsound wheat grains caused due to the imaging system and background noise. In this paper, a CBDNet-V terahertz spectral image enhancement model is proposed. Compared with the traditional algorithms, the peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) of the obtained enhanced images using the proposed model show performance improvement. As validated by the ResNet-50 classification network, the proposed model processes images with an accuracy of 94.8%, and the recognition accuracy is improved by 3.7% and 1.9%, respectively, compared to the images with only denoising and feature extraction. The experimental results indicate that the deep learning-based terahertz spectral image technology for unsound wheat kernels has good prospects in the identification of unsound wheat kernels.

Agricultural products need to be inspected for quality and safety, and the issue of safety of agricultural products caused by quality is frequently investigated. Safety testing should be carried out before agricultural products are consumed. The existing technologies for inspecting agricultural products are time-consuming and require complex operation, and there is motivation to develop a rapid, safe, and non-destructive inspection technology. In recent years, with the continuous progress of THz technology, THz spectral imaging, with the advantages of its unique characteristics, such as low energies, superior spatial resolution, and high sensitivity to water, has been recognized as an efficient and feasible identification tool, which has been widely used for the qualitative and quantitative analyses of agricultural production. In this paper, the current main performance achievements of the use of THz images are presented. In addition, recent advances in the application of THz spectral imaging technology for inspection of agricultural products are reviewed, including internal component detection, seed classification, pesticide residues detection, and foreign body and packaging inspection. Furthermore, machine learning methods applied in THz spectral imaging are discussed. Finally, the existing problems of THz spectral imaging technology are analyzed, and future research directions for THz spectral imaging technology are proposed. Recent rapid development of THz spectral imaging has demonstrated the advantages of THz radiation and its potential application in agricultural products. The rapid development of THz spectroscopic imaging combined with deep learning can be expected to have great potential for widespread application in the fields of agriculture and food engineering.

Terahertz (THz) spectroscopy and multivariate data analysis were explored to discriminate eight wheat varieties. The absorption spectra were measured using THz time-domain spectroscopy from 0.2 to 2.0 THz. Using partial least squares (PLS), a regression model for discriminating wheat varieties was developed. The coefficient of correlation in cross validation (R) and root-mean-square error of cross validation (RMSECV) were 0.985 and 1.162, respectively. In addition, interval PLS was applied to optimize the models by selecting the most appropriate regions in the spectra, improving the prediction accuracy (R = 0.992 and RMSECV = 0.967). Results demonstrate that THz spectroscopy combined with multivariate analysis can provide rapid, nondestructive discrimination of wheat varieties.

In order to improve the detection accuracy for the quality of wheat, a recognition method for wheat quality using the terahertz (THz) spectrum and multi-source information fusion technology is proposed. Through a combination of the absorption and the refractive index spectra of samples of normal, germinated, moldy, and worm-eaten wheat, support vector machine (SVM) and Dempster-Shafer (DS) evidence theory with different kernel functions were used to establish a classification fusion model for the multiple optical indexes of wheat. The results showed that the recognition rate of the fusion model for wheat samples can be as high as 96%. Furthermore, this approach was compared to the regression model based on single-spectrum analysis. The results indicate that the average recognition rates of fusion models for wheat can reach 90%, and the recognition rate of the SVM radial basis function (SVM-RBF) fusion model can reach 97.5%. The preliminary results indicated that THz-TDS combined with DS evidence theory analysis was suitable for the determination of the wheat quality with better detection accuracy.