Mosquitoes are one of the deadliest insects, causing harm to humans worldwide. Preemptive prevention and forecasting are important to prevent mosquito-borne diseases. However, current mosquito identification is mostly conducted manually, which consumes time, wastes labor, and causes human error. In this study, we developed an automatic image analysis method to identify mosquito species using a deep learning-based object detection technique. Color and fluorescence images of live mosquitoes were acquired using a mosquito capture device and were used to develop a deep learning-based object detection model. Among the deep learning-based object identification models, the combination of a swine transformer and a faster region-convolutional neural network model demonstrated the best performance, with a 91.7% F1-score. This indicates that the proposed automatic identification method can be rapidly applied for efficient analysis of species and populations of vector-borne mosquitoes with reduced labor in the field.

This study proposes a deep-learning-based system for detecting and classifying Scirtothrips dorsalis Hood, a highly invasive insect pest that causes significant economic losses to fruit crops worldwide. The system uses yellow sticky traps and a deep learning model to detect the presence of thrips in real time, allowing farmers to take prompt action to prevent the spread of the pest. To achieve this, several deep learning models are evaluated, including YOLOv5, Faster R-CNN, SSD MobileNetV2, and EfficientDet-D0. EfficientDet-D0 was integrated into the proposed smartphone application for mobility and usage in the absence of Internet coverage because of its smaller model size, fast inference time, and reasonable performance on the relevant dataset. This model was tested on two datasets, in which thrips and non-thrips insects were captured under different lighting conditions. The system installation took up 13.5 MB of the device’s internal memory and achieved an inference time of 76 ms with an accuracy of 93.3%. Additionally, this study investigated the impact of lighting conditions on the performance of the model, which led to the development of a transmittance lighting setup to improve the accuracy of the detection system. The proposed system is a cost-effective and efficient alternative to traditional detection methods and provides significant benefits to fruit farmers and the related ecosystem.

Ensuring food safety in fresh-cut vegetables is essential due to the frequent presence of foreign material (FM) that threatens consumer health and product quality. This study presents a real-time FM detection system developed using the YOLO object detection framework to accurately identify diverse FM types in cabbage and green onions. A custom dataset of 14 FM categories—covering various shapes, sizes, and colors—was used to train six YOLO variants. Among them, YOLOv7x demonstrated the highest overall accuracy, effectively detecting challenging objects such as transparent plastic, small stones, and insects. The system, integrated with a conveyor-based inspection setup and a Python graphical interface, maintained stable and high detection accuracy confirming its robustness for real-time inspection. These results validate the developed system as an alternative intelligent quality-control layer for continuous, automated inspection in fresh-cut vegetable processing lines, and establish a solid foundation for future robotic-based removal systems aimed at fully automated food safety assurance.

김 화 수 1 •이 상 준 2 대구대학교 언어치료학과 1 , 대구대학교 일반대학원 재활과학과

Optical seed-vigor phenotyping enables rapid screening by near-infrared spectroscopy, hyperspectral imaging, Raman spectroscopy, and related methods. These measurements can predict seed performance without identifying the biochemical processes underlying a prediction. This narrative review examines how surface-enhanced Raman scattering (SERS) could add targeted molecular information to optical phenotyping. It distinguishes spontaneous Raman measurements, SERS biomarker assays, seed-matrix analytical validation, and biological outcome linkage. The proposed application uses extracts from representative seed subsets; its destructive character arises from sample preparation, whereas Raman and SERS measurements can be non-destructive under suitable conditions. Seed Raman studies already support useful classification and relative chemical measurements. Among the ten biomarker-sensor papers mapped here, however, none combines seed-matrix calibration and reference chemistry with matched vigor outcomes and independent lot testing. This finding is limited to the mapped corpus rather than to all published research. Protein adsorption, lipid-rich co-extractives, pigments, and substrate variability require crop-specific assessment before quantitative SERS results can be transferred between seed types. A feasible starting point is a small target panel, paired extract aliquots for SERS and reference analysis, and germination testing of matched seed subsets. Studies in other matrices guide this development but do not establish seed-vigor performance.

The quality of meat emulsions during mixing is difficult to control because rapid, objective, and non-destructive measurement methods remain limited. This study investigated the effects of lean meat-to-fat ratio and mixing time on moisture content in meat batters using short-wave infrared (1,000-1,700 nm) hyperspectral imaging combined with chemometric modeling. Meat batters were prepared at three lean meat-to-fat ratios: Group A (6:2), Group B (5:3), and Group C (4:4), and mixed for 0, 1, 2, 5, 7, 10, and 15 min. Moisture content differed significantly among formulations, with Group A showing the highest value (63.4%), followed by Group B (57.2%) and Group C (52.4%). Mixing time also had a significant effect, with the highest moisture content observed at 0 min (61.8%) and stabilization at approximately 56–58% after 1–15 min, indicating progressive homogenization. Two-way ANOVA revealed significant main effects of lean meat-to-fat ratio and mixing time (p < 0.001), but no significant interaction. Important wavelengths were identified at 1,000-1,124, 1,190-1,230, near 1,400, and 1,600-1,700 nm, mainly related to water, fat, and protein absorptions. PLSR and Elastic Net emphasized lower water-related wavelengths, whereas RF highlighted mid- and high-wavelength regions associated with fat and protein. Among the tested models, Elastic Net with raw spectra showed the best prediction performance (R²P = 0.76, RMSEP = 2.54%), while the ensemble model also showed strong performance. In addition, hyperspectral imaging visualized moisture distribution and reduced sample heterogeneity with increased mixing time. These results suggest that SWIR hyperspectral imaging is an effective non-destructive tool for predicting moisture content and monitoring mixing uniformity in meat batters. Graphical Abstract