Aviation Industry Corporation of China (China)

Semantic segmentation of moving objects in dynamic backgrounds faces core challenges such as background interference and blurred target features. This study proposes an innovative architecture that integrates Generative Adversarial Network (GAN) with Transformers. The GAN module enhances adaptability to dynamic backgrounds through adversarial training, while the self-attention mechanism in the Transformer captures long-range semantic dependencies. A gated fusion strategy is designed to achieve dynamic balancing of multimodal features. The method employs a conditional GAN to generate dynamic background samples with variations in illumination and motion blur. A Transformer-based encoder-decoder structure is used to model global contextual relationships. A temporal attention module is introduced to incorporate motion vector fields, improving temporal consistency. Additionally, a KL-divergence (KL) constrained semantic consistency loss optimizes the plausibility of generated samples. Experiments are conducted on both a multi-dimensional simulated dataset and the real-world KITTI dataset. Results show that the proposed model achieves an average Intersection over Union (IoU) of 85.6% in standard dynamic scenes, outperforming DeepLabv3 + by 9.2% points. In low-light and high-speed motion scenarios, the robustness index reaches 92.0%, 8.5 points higher than baseline models. Ablation studies demonstrate that removing the Transformer leads to a 6.7% drop in mIoU, while excluding the feature fusion module reduces robustness by 4.0%, confirming the necessity of both components. Temporal analysis reveals that the model maintains a stable performance of 84.5–86.5% over 20-frame sequences, with fluctuation reduced by 63% compared to baseline. The adversarial training improves the model’s adaptability to lighting changes by 5.3%. The multi-head self-attention (MSA) mechanism reduces long-range misclassification by 6.7%. The gated fusion strategy lowers false positive rates in background-disturbed regions by 12.8%. This framework optimizes segmentation through a generator-segmenter feedback loop, effectively balancing dynamic background noise suppression and semantic fidelity. The contributions are threefold: (1) The first semantic segmentation framework to deeply integrate GANs and Transformers. (2) A theoretical model for dynamic feature gating and semantic consistency constraints. (3) A standardized evaluation system covering 10 dynamic background types and five illumination gradients. This study provides key technical support for real-time environmental perception in autonomous driving and intelligent surveillance, advancing both the theoretical and practical frontiers of dynamic scene understanding.

Abstract Objective To examine the independent effects of the serum creatinine to cystatin C ratio (CCR) and depression on the risk of disability, as well as their interaction, among older adults in China. Methods This study utilized data from the 2015 China Health and Retirement Longitudinal Study (CHARLS) and included 5,192 individuals aged ≥ 60 years. Multivariable logistic regression models were employed to analyze the independent associations of CCR and depression with disability, and multiplicative interaction terms were introduced to evaluate their interaction. Additionally, restricted cubic splines were used to explore the dose-response relationship between CCR and the risk of disability. Results Among all participants, the prevalence of disability was 22.65% (1,176/5,192). After multivariable adjustment, both a low CCR level (OR = 1.20, 95% CI: 1.02–1.42, P = 0.025) and depression (OR = 2.13, 95% CI: 1.84–2.48, P < 0.001) were independent risk factors for disability. Interaction analysis revealed a significant multiplicative interaction effect between CCR and depression (OR = 2.695, 95% CI: 2.150–3.386), indicating that their combined effect on disability risk exceeded the product of their individual independent effects. Restricted cubic spline analysis suggested a nonlinear negative correlation between CCR and the risk of disability (P for nonlinearity = 0.033). Conclusion In the Chinese older adult population, both low-level CCR and depression are significant independent risk factors for disability. Moreover, they exhibit a synergistic interaction, jointly exacerbating the risk of disability. It is recommended that clinical management and public health interventions for disability in older adults adopt integrated strategies that combine muscle function preservation, physical function rehabilitation, and mental health promotion.

Edge cracking severely limits the rolling yield of magnesium alloy plates. A novel lattice severe deformation rolling (LSDR) process using corrugated rolls is proposed to suppress edge cracking. Numerical simulations, rolling experiments, and microstructural analyses were conducted, with results compared to conventional flat rolling (FR), to elucidate the suppression mechanism. LSDR induces a multi-peak stress distribution and restricts metal flow, thereby reducing additional stresses responsible for edge cracking. Deformation heat generated in local severe deformation zones compensates for thermal loss, alleviates the temperature gradient between the plate edge and center, and enhances overall plasticity. According to the Cockcroft–Latham fracture criterion, LSDR effectively limits damage growth and confines damage within a single lattice, suppressing crack propagation, whereas FR produces damage values far exceeding the critical value of 0.43. Furthermore, fine grains formed in severe deformation zones, together with dislocation entanglement induced by twinning, impede crack propagation. This work demonstrates the effectiveness of LSDR and provides a new approach for mitigating edge cracking in rolled metal plates.

Diabetic retinopathy (DR) is a critical and prevalent microvascular complication in patients with diabetes mellitus (DM), potentially culminating in blindness. Although lncRNA HCG11 has been proposed as a biomarker for this type of disease, supporting evidence remains sparse. This research endeavor is aimed at explicating the clinical relevance and molecular underpinnings of lncRNA HCG11 in DR patients. Serum levels of HCG11 were quantified in both DR and DM patients to assess its potential as an early diagnostic biomarker. Additionally, the effects of HCG11 on human retinal pigment epithelial cells (ARPE-19) were investigated in a vitro cell model, and the interaction between HCG11 and miR-532-3p was investigated through a dual-luciferase reporter assay. The results indicated that a significant upregulation of HCG11 in DR patients and could distinguish DR patients from those with DM with high sensitivity and specificity. In vitro experiments revealed that knockdown of HCG11 mitigated high glucose-induced inhibition of cell viability, reduced apoptosis, and lowered inflammatory cytokine secretion. The dual-luciferase reporter assay confirmed the binding interaction between HCG11 and miR-532-3p, also noting a negative correlation. The detailed mechanisms of their interaction warrant further investigation. lncRNA HCG11 was pivotal in the pathogenesis of DR, emerging as a promising diagnostic and therapeutic target. This study offers novel molecular insights and potential strategies for diagnosing and treating diabetic retinopathy.

Channel estimation (CE) is essential for attaining the benefits of carrier aggregation (CA), such as transmission rate and system capacity. However, most existing methods exhibit high resource consumption and relatively low CE accuracy, seriously hindering their practical application. To address these issues, a sensing and channel correlation-aided CE method is proposed in this paper. Specifically, we utilize the sensed delay and Doppler spreads to enhance the CE accuracy of the primary cell (PCell). With the enhanced CE of PCell, we extract path prior information for the CE of secondary cells (SCells). By leveraging this path prior information and the channel correlation between PCell and SCells, an initial reconstruction of the SCells channel is performed. With this initial reconstruction, an iterative channel enhancement scheme is further developed to address the issue of weak path correlation. Simulation results validate the effectiveness of the proposed method and its robustness against parameter variations.

Automatic modulation recognition (AMR) is an essential technology for spectrum sensing and management. Machine learning-based AMR techniques have garnered significant research interest. However, in practical applications, the effectiveness of AMR is hindered by variations in data distribution due to changes in carrier frequency, sample rate, and channel model. Existing deep learning methods struggle to adapt to these diverse distributions, and training separate datasets for each scenario is often impractical. To address these challenges, we propose a domain adaptive hybrid network (DAHNet) that integrates convolutional neural networks (CNNs) and long short-term memory (LSTM) networks. This architecture leverages the phase and amplitude of constellation points, as well as time-series data from I/Q sequences. Additionally, by explicitly minimizing the maximum mean discrepancy (MMD) between features from the source and target domains across multiple neural layers, we tackle the model generalization difficulties posed by varying propagation conditions and noise in complex electromagnetic environments. We validate our method using deep transfer learning on real-world data, with training results derived from the publicly available RadioML2016.10a and RadioML2018.01a datasets. Experimental results demonstrate the proposed method’s superior generalization ability across different spectrum scenarios.