Yakai He

Ensuring the surface quality of fresh soybeans is critical for maintaining their commercial value and consumer confidence. However, traditional manual inspection remains labor-intensive, subjective, and inadequate for real-time, high-throughput sorting. In this study, we present a multi-scale attention-augmented You Only Look Once version 8 (YOLOv8) framework tailored for real-time surface defect detection in fresh soybeans. The proposed model integrates two complementary attention mechanisms—Squeeze-and-Excitation (SE) and Multi-Scale Dilated Attention (MSDA)—to enhance the detection of small, irregular, and low-contrast defects under complex backgrounds. Rather than relying on cross-model comparisons, we perform systematic ablation studies to evaluate the individual and combined contributions of SE and MSDA across diverse defect categories. Experimental results from a custom-labeled soybean dataset demonstrate that the integrated SE+MSDA model achieves superior performance in terms of precision, recall, and Mean Average Precision (mAP), particularly for challenging categories such as wormholes and speckles. The proposed framework provides a lightweight, interpretable, and deployment-ready solution for intelligent agricultural inspection, with potential applicability to broader food quality control tasks.

This study investigated the optical properties (OPs) and Monte Carlo (MC) simulations of light propagation in Healthy Group (HG) and Blackhearted Group (BG) potatoes. The MC simulation of light propagation indicated that both the photon packet weight and the penetration depth were significantly lower in blackhearted tissues than in healthy tissues. The simulation revealed deeper light penetration in healthy tissues than in the blackhearted tissues, approximately 6.73 mm at 805 nm, whereas the penetration depth in blackhearted tissues was much shallower (1.30 mm at 805 nm). Additionally, the simulated absorption energy at both 490 nm and 805 nm was higher in blackhearted tissues, suggesting that these wavelengths effectively detect blackheart in potatoes. The absorption (μ_a) and reduced scattering (μ’_s) coefficients were obtained using Vis-NIR spectroscopy, which represented a notable increase in μ_a in BH tissues, particularly around 550–850 nm, and an increase in μ’_s across the Vis-NIR region. Based on transmittance (Tt), μ_a and μ’_s, Support Vector Machine Discriminant Analysis (SVM-DA) models demonstrated exceptional performance, achieving 95.83–100.00% accuracy in Cross-Validation sets, thereby confirming the robustness and reliability of the optical features for accurate blackheart detection. These findings provide valuable theoretical insights into the accuracy and robustness of predictive models for detecting blackhearted potatoes.

More efficient resource utilization and increased crop utilization rate are needed to address the growing demand for food. The efficient quality testing of key agricultural products such as potatoes, especially the rapid testing of key nutritional indicators, has become an important strategy for ensuring their quality and safety. In this study, visible and near infrared (Vis/NIR) transmittance spectroscopy (600–900 nm) was used for the online analysis of multiple quality parameters in potatoes. The study concentrated on comparing three one-dimensional convolutional neural network (1D-CNN) models, specifically, the fine-tuned DeepSpectra, the fine-tuned 1D-AlexNet, and classic CNN, with UVE-PLS (uninformative variable elimination–partial least squares) models. These models utilized spectral data for the real-time detection of dry matter (DM) content in potatoes. To address the challenges posed by limited data from Vis/NIR, this study strategically implemented data augmentation techniques. This approach significantly enhanced the robustness and generalization capabilities of the models. The 1D-AlexNet and DeepSpectra models achieved 0.934 and 0.913 R²_P and 0.0603 and 0.0695 g/100 g RMSEP for DM, respectively. Compared to UVE-PLS, the R²_P value improved by 21.31% (0.770 to 0.934) for the 1D-AlexNet model and 18.64% (0.770 to 0.913) for the DeepSpectra model. The RMSEP value was reduced by 47.31% (0.114 to 0.0603) for 1D-AlexNet, and 39.30% (0.114 to 0.0695) for the DeepSpectra model. As a result, this study would be helpful for researching the online Vis/NIR transmission determination of potato DM using deep learning. These results highlighted the immense potential of employing specific spectral features in deep-learning models for a more precise and efficient online assessment of agricultural quality. This advancement provided some insight and reference for further contributing to the evolution of more targeted and efficient quality assessment methods in agricultural products.

Spraying pesticides using row alignment in the maize seedling stage can effectively improve pesticide utilization and protect the ecological environment. Therefore, this study extracts a guidance line for maize crops using machine vision and develops an automatic row-oriented control system based on a high-clearance sprayer. First, the feature points of crop rows are extracted using a vertical projection method. Second, the candidate crop rows are obtained using a Hough transform, and two auxiliary line extraction methods for crop rows based on the slope feature outlier algorithm are proposed. Then, the guidance line of the crop rows is fitted using a tangent formula. To greatly improve the robustness of the vision algorithm, a Kalman filter is used to estimate and optimize the guidance line to obtain the guidance parameters. Finally, a visual row-oriented spraying platform based on autonomous navigation is built, and the row alignment accuracy and spraying performance are tested. The experimental results showed that, when autonomous navigation is turned on, the average algorithm time consumption of guidance line detection is 42 ms, the optimal recognition accuracy is 93.3%, the average deviation error of simulated crop rows is 3.2 cm and that of field crop rows is 4.36 cm. The test results meet the requirements of an automatic row-oriented control system, and it was found that the accuracy of row alignment decreased with increasing vehicle speed. The innovative spray performance test found that compared with the traditional spray, the inter-row pesticide savings were 20.4% and 11.4% overall, and the application performance was significantly improved.

Planting uniformity performance of motor-driven maize precision seeding systems