北京工业职业技术学院

CrSb, as a prototypical candidate compound of altermagnetism, has attracted significant attention due to its unique magnetic properties. In this study, we combine systematic electrical transport measurements with first-principles calculations to investigate the possible realization mechanisms of topological semimetal states in CrSb and their manifestations in quantum transport phenomena. Our high-field magneto-transport measurements reveal that the magnetoresistance of CrSb exhibits no sign of saturation up to 35 T, following a distinct power-law dependence with an exponent of 1.48. The nonlinear Hall resistivity curve further indicates a multiband charge transport mechanism. Under high magnetic fields, we observe pronounced Shubnikov-de Haas (SdH) quantum oscillations and discernible Zeeman-effect-induced band splitting at 1.6 K. Systematic Fermi surface and band calculations combined with Berry phase analysis confirm the nontrivial topological character of this material (with a Berry phase approaching π). These findings not only provide crucial experimental evidence for understanding the electronic structure of CrSb, but also establish an important foundation for investigating topological quantum states in altermagnets.

Building extraction underpins land-use assessment, urban planning, and disaster mitigation, yet dense urban scenes still cause missed small objects, target adhesion, and ragged contours. We present High-Resolution-Mamba (HR-Mamba), a high-resolution semantic segmentation network that augments a High-Resolution Network (HRNet) parallel backbone with edge-aware and sequence-state modeling. A Canny-enhanced, median-filtered stem stabilizes boundaries under noise; Involution-based residual blocks capture position-specific local geometry; and a Mamba-based State Space Models (Mamba-SSM) global branch captures cross-scale long-range dependencies with linear complexity. Training uses a composite loss of binary cross entropy (BCE), Dice loss, and Boundary loss, with weights selected by joint grid search. We further design a feature-driven adaptive post-processing pipeline that includes geometric feature analysis, multi-strategy simplification, multi-directional regularization, and topological consistency verification to produce regular, smooth, engineering-ready building outlines. On dense urban imagery, HR-Mamba improves F1-score from 80.95% to 83.93%, an absolute increase of 2.98% relative to HRNet. We conclude that HR-Mamba jointly enhances detail fidelity and global consistency and offers a generalizable route for high-resolution building extraction in remote sensing.

To address low grading efficiency, delayed feedback, and insufficient personalized guidance in higher vocational workplace English writing, this study constructs a deep learning-driven intelligent marking system and verifies its teaching effect through empirical research. A total of 120 students from two higher vocational English classes were divided into an experimental group using the intelligent grading system and a control group using conventional teacher grading. During the 16-week experiment, writing tests, questionnaires, interviews, and learning records were collected and statistically analyzed. The system integrates BERT, LSTM, natural language processing, and workplace-document evaluation rules to assess grammatical accuracy, sentence diversity, content relevance, and workplace standardization. The results show that the experimental group significantly outperformed the control group in overall writing scores and all core dimensions, while writing interest, autonomous learning, and confidence also improved. The proposed system reduces teacher workload and provides a data-driven feedback pipeline suitable for smart educational environments based on networked human-machine interaction.

Image aesthetic multi-attribute captioning emphasizes fine-grained aesthetic attributes, capturing intricate aesthetic characteristics from diverse perspectives and reflecting a more nuanced and profound understanding of aesthetics, encompassing a wide spectrum of aesthetic semantics. Despite its potential, current approaches to aesthetic multi-attribute captioning remain underexplored. This paper introduces a novel image aesthetic multi-attribute captioning method grounded in vision-language pre-training, aimed at addressing the inadequacy in aesthetic information expression by generating fine-grained attribute-aware aesthetic descriptions to enrich semantic depth and interpretability. Adopting a “pre-training and fine-tuning” paradigm, the proposed framework leverages CLIP and GPT-2 architectures, aligning CLIP-derived visual features with the GPT-2 embedding space via a cross-modal mapping network. The incorporation of aesthetic attribute control flags enable precise regulation of the generated aesthetic multi-attribute captions. Experimental results demonstrate that our method surpasses mainstream approaches across several metrics on DPC-MAC and PCCD dataset, including BLEU, METEOR, SPICE, etc. Furthermore, ablation studies on multi-stage aesthetic pre-training substantiate the effectiveness of the proposed strategy. The model consistently produces aesthetically coherent and attribute-aligned captions, underscoring its potential for advanced aesthetic analysis.

Addressing the critical challenges of insufficient precision in landslide hazard identification and the lack of dynamic stability assessment for high-steep slopes in open-pit mines, this study innovatively proposes an integrated technical framework that deeply fuses time-series Interferometric Synthetic Aperture Radar (InSAR) with machine learning (ML), achieving an intelligent analysis framework that integrates deformation monitoring and stability assessment for open-pit mine landslide hazards. The main contributions include: (1) overcoming the limitations of InSAR technology in low-coherence areas, an improved Small Baseline Subset InSAR (SBAS-InSAR) algorithm was adopted to extract slope deformation, increasing the monitoring point density on complex rock slopes by a factor of 2.18, obtaining high-precision deformation fields, and significantly enhancing the deformation capture capability of high-steep slopes; (2) a new paradigm of dynamic-static multi-factor coupled stability assessment was proposed, which deeply fuses time-series InSAR deformation characteristics with multi-source heterogeneous data, including geological, mining, and environmental factors, employing a dual-model collaborative strategy of Random Forest (RF) and XGBoost achieving an Area Under the Curve (AUC) exceeding 0.88, with the InSAR dynamic factor contributing the highest importance, thereby validating the core role of dynamic monitoring data in stability assessment. The empirical study at a large open-pit mine in northern China demonstrates that high- and very-high-risk zones are precisely localized to specific benches, providing an operational technical support system for mine safety production and offering significant demonstration value for promoting the deep application of InSAR technology in the field of mine disaster early warning.