In a printing production process, paper often exhibits various surface defects due to objective factors. These defects manifest in different forms and are typically small and densely distributed, posing a challenge for existing object detection methods. To tackle this issue, this research reports a paper product appearance inspection system that combines a dual-layer template and an attention mechanism. The system effectively addresses the problem of surface detect detection by matching template images with the images to be inspected. This approach ensures robust detection performance and reduces false negatives through the utilization of the improved You Only Look Once v7 algorithm. The enhanced algorithm incorporates diverse attention mechanisms that effectively utilize features, thereby enhancing the model’s capability for defect detection. The system’s development includes the creation of a visual interface that provides a customized defect detection solution for the printing authority. Through comprehensive ablation experiments, the algorithm demonstrates significant improvements in both detection accuracy and speed. This system enables automatic inspection of the appearance of single-page paper products and automates the removal of products with surface defects. Consequently, it achieves a high level of automation in the inspection and handling processes for single-page paper products.

Spirometry plays a key role in diagnosing respiratory diseases, but its accuracy often falls short of clinical expectations. While deep learning models have shown promise in automating spirometry analysis, challenges persist. Spirometry curves are vulnerable to noise from minor signals, which can lead to diagnostic errors. Moreover, the high computational demands of current algorithms limit their use in clinical practice. To address these issues, we present a novel approach that integrates n-Order Adaptive Fourier Decomposition with deep learning techniques to enhances quality control in spirometry analysis. Adaptive Fourier Decomposition improves the resolution and processing of flow-volume curves, effectively minimizing noise. By leveraging the strengths of deep learning models alongside AFD, our method accurately detects small and complex abnormalities in spirometry data. Ablation studies show that our method outperforms traditional approaches, raising the mean average precision from 89.5 percent to 95.5 percent. Furthermore, the model’s lightweight design, achieved through computational optimizations and structural simplifications, enables efficient deployment in various clinical settings, improving diagnostic accuracy and accessibility.

Current paper defect detection methods often struggle to satisfy the critical demands for precision and adaptability in complex industrial environments. This study developed a novel multi-sensor platform combining low-precision LiDAR and high-resolution camera, along with an Adaptive Sensor Fusion Module - Adaptive Fourier Decomposition (ASFM-AFD) framework for processing multi-modal data. This integration substantially improves defect detection capabilities. The AFD-based optimization approach effectively enhances the quality of low-precision LiDAR data, achieving performance comparable to high-precision LiDAR while providing cost-effective solution for industrial inspection applications. Comparative studies with Empirical Mode Decomposition and Variational Mode Decomposition demonstrates that AFD exhibits superior noise reduction and resolution enhancement while maintaining stable performance under various challenging conditions. Experimental results show that the proposed framework achieves an average precision of 97.5% for paper defect detection in high-speed production environments, significantly improving detection accuracy and demonstrating robustness across various operating conditions.

This study employs an unsupervised learning technique to identify features of misinformation on Weibo, a popular social media platform in China. By utilizing BERT (Bidirectional Encoder Representations from Transformers) for sequence classification, we were able to detect a dataset with a mix of true and false information, achieving a perfect recall rate for false information detection. However, the precision for true information was low, with a 3% precision, 6% recall, and an F1 score of 6%, indicating a high rate of misclassification. To address this, we conducted a Latent Dirichlet Allocation (LDA) analysis on the misclassified true information, identifying specific features that led to incorrect classification. Additionally, social network analysis revealed the presence of structural holes within the information network. This study contributes to the understanding of misinformation detection mechanisms and provides insights into the social dynamics of information spread.

Bougainvillea glabra, commonly planted landscape flowers in Macau, requires the precise water management during whole growing process. Conventional timed irrigation systems, however, fail to offer the required level of precision. Sensor-based watering systems have been employed, utilizing soil moisture and temperature sensors to monitor plant conditions. Nonetheless, these sensors are proved to be unreliable due to various soil factors, such as the presence of stones. Thus, plants still require botanical experts to judge the status based on plant photos. To address this challenge and provide a dependable evaluation of the plant’s status, this study proposes a novel approach: a Plant Status Detection Net (PSDNet) for Bougainvillea sensing, which uses computer vision technology to replace botanical experts in decision-making. Different from directly using image classification, this paper combines object detection to extract plant leaves, and then makes decisions based on leaves. By utilizing the Regions of Interest (RoI) structure from a pre-trained module called Leaf Regions with CNN Features (Leaf-RCNN), the proposed PSDNet effectively extracts leaf regions and their corresponding feature maps from captured images. To further improve accuracy, a specialized decoupling head and a position embedding module are integrated into the network to enable the extraction of relative position information between leaves. Finally, in a pilot project at Macau Keang Peng Middle School (MKPMS), an automated watering system was implemented using PSDNet, leading to a substantial increase in the flowering rate. The anticipated adoption of this watering system is projected to significantly advance landscaping practices in Macau and beyond.

This paper presents an efficient grasp pose estimation algorithm for robotic arm systems with a two-finger parallel gripper and a consumer-grade depth camera. Unlike traditional deep learning methods, which suffer from high data dependency and inefficiency with low-precision point clouds, the proposed approach uses ellipsoidal modeling to overcome these issues. The algorithm segments the target and then applies a three-stage optimization to refine the grasping path. Initial estimation fits an ellipsoid to determine principal axes, followed by nonlinear optimization for a six-degree-of-freedom grasp pose. Validation through simulations and experiments showed a target grasp success rate (TGSR) of over 83% under low noise, with only a 4.9% drop under high noise—representing a 68.0% and a 42.4% improvement over GPD and PointNetGPD, respectively. In real-world tests, success rates ranged from 95 to 100%, and the computational efficiency was improved by 56.3% compared to deep learning methods, proving its practicality for real-time applications. These results demonstrate stable and reliable grasping performance, even in noisy environments and with low-cost sensors.