Jun Zhang’s research while affiliated with Nankai University and other places

RGB-D salient object detection (SOD), aiming to highlight prominent regions of a given scene by jointly modeling RGB and depth information, is one of the challenging pixel-level prediction tasks. Recently, the dual-attention mechanism has been devoted to this area due to its ability to strengthen the detection process. However, most existing methods directly fuse attentional cross-modality features under a manual-mandatory fusion paradigm without considering the inherent discrepancy between the RGB and depth, which may lead to a reduction in performance. Moreover, the long-range dependencies derived from global and local information make it difficult to leverage a unified efficient fusion strategy. Hence, in this paper, we propose the GL-DMNet, a novel dual mutual learning network with global-local awareness. Specifically, we present a position mutual fusion module and a channel mutual fusion module to exploit the interdependencies among different modalities in spatial and channel dimensions. Besides, we adopt an efficient decoder based on cascade transformer-infused reconstruction to integrate multi-level fusion features jointly. Extensive experiments on six benchmark datasets demonstrate that our proposed GL-DMNet performs better than 24 RGB-D SOD methods. Codes and results are available at https://github.com/kingkung2016/GL-DMNet.

Root systems are pivotal for nutrient absorption, exhibiting high plasticity in phosphorus (P) acquisition, and significantly influencing soil phosphorus availability. However, the impacts of different P application methods on root parameters and P utilization efficiency in cotton (Gossypium hirsutum L.) under Xinjiang conditions are still not well understood. To identify optimal P fertilization strategies, a consecutive two-year field experiment (2023–2024) under mulched drip irrigation was conducted. Three P application methods were tested: no P (CK), basal P application (PB), and drip P application (PD). Results revealed that P application methods significantly affected cotton dry matter, P use efficiency, root morphology, and yield (p < 0.05). Over the two years, the optimized treatment (25% P applied at bud stage and 25% at flowering-boll stage, PD) increased yield by 13.62% and 9.50% compared to full basal application (PB), with P use efficiency improved by 22.04–31.51% and agronomic efficiency improved by 6.56–9.75 kg kg−1. PB significantly increased soil-available P in 0–20 cm (34.17–70.09%) and 20–40 cm layers (30.37–70.32%) compared to CK. During the bud stage, PD treatment exhibited higher soil-available P in the 20–40 cm layer than PB. PD enhanced P uptake and dry matter accumulation, with increases of 22.43–36.33% and 7.90–15.55% in reproductive organ P accumulation compared to other treatments. Root parameters followed PD > PB > CK across all treatments. At the seedling stage, PB increased total root length by 19.79% compared to CK, while PD increased root volume by 46.15% compared to PB. During the bud stage, PB increased root volume by 53.33% compared to CK, and PD enhanced root surface area and volume by 39.25% and 47.82% compared to PB. Root volume showed a significant positive correlation with phosphorus absorption across growth stages. The PD treatment significantly enhanced soil P availability and P use efficiency and optimized root spatial distribution. This treatment consistently increased cotton yield by 30.41–39.09% (p < 0.05) compared to CK, demonstrating stable positive effects. This study highlights that adjusting P application methods can establish sustainable, high-yield agricultural fertilization systems.

Purpose To explore left atrial (LA) function in patients with systemic lupus erythematosus (SLE) and mitral regurgitation assessed by cardiac magnetic resonance imaging feature tracking (MRI-FT). Method In this study, 64 SLE patients (35.0 ± 12.3 years, 49 females) and 50 age- and gender-matched healthy controls (32.1 ± 11.2 years, 32 females) were analyzed. Patients with SLE were further classified into two subgroups according to mitral regurgitation defined by echocardiography: mitral regurgitation subgroup ( n = 32) and non-mitral regurgitation subgroup ( n = 32). All subjects underwent cardiac magnetic resonance (CMR) examination and SLE patients underwent extra laboratory testing. LA volume and strain parameters were compared among the three groups, and a correlation analysis was further performed between CMR parameters and clinical variables. Results Patients in both mitral regurgitation and non-mitral regurgitation subgroups showed higher LAVmin (29.15 ± 7.5, 26.22 ± 8.23 vs 21.68 ± 7.67, p = .003, .026) and LAVimin (17.44,14.9 vs 12.62, p = .002,.046) than healthy control subjects. Abnormal LA reservoir, conduit and bump strain were also observed in SLE patients (all p < .05), with more severe reservoir ( p = .006, .031) and bump strain ( p = .01, .033) abnormality found in mitral regurgitation subgroup. In the SLE group, LA reservoir, bump and conduit strain were significantly positively correlated with LVEF (B = 0.467, p < .001 vs B = 0.019, p = .01 vs B = 0.168, p = .001 vs B = 0.036, p < .001 vs B = 0.020, p = .024). Conduit strain (εe, SRe) was positively correlated with LAEF (B = 0.229, p = .032 vs B = 0.027, p = .008). Conclusion MRI-FT based LA parameters may be considered a helpful tool to evaluate LA function in SLE patients, and LA strain may indicate severity of the cardiac dysfunction in SLE patients.

Mobile crowdsensing is evolving beyond traditional human-centric models by integrating heterogeneous entities like unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs). Optimizing task allocation among these diverse agents is critical, particularly in challenging emergency rescue scenarios characterized by complex environments, limited communication, and partial observability. This paper tackles the Heterogeneous-Entity Collaborative-Sensing Task Allocation (HECTA) problem specifically for emergency rescue, considering humans, UAVs, and UGVs. We introduce a novel ``Hard-Cooperative'' policy where UGVs prioritize recharging low-battery UAVs, alongside performing their sensing tasks. The primary objective is maximizing the task completion rate (TCR) under strict time constraints. We rigorously formulate this NP-hard problem as a decentralized partially observable Markov decision process (Dec-POMDP) to effectively handle sequential decision-making under uncertainty. To solve this, we propose HECTA4ER, a novel multi-agent reinforcement learning algorithm built upon a Centralized Training with Decentralized Execution architecture. HECTA4ER incorporates tailored designs, including specialized modules for complex feature extraction, utilization of action-observation history via hidden states, and a mixing network integrating global and local information, specifically addressing the challenges of partial observability. Furthermore, theoretical analysis confirms the algorithm's convergence properties. Extensive simulations demonstrate that HECTA4ER significantly outperforms baseline algorithms, achieving an average 18.42% increase in TCR. Crucially, a real-world case study validates the algorithm's effectiveness and robustness in dynamic sensing scenarios, highlighting its strong potential for practical application in emergency response.

This study explores the challenges of transient stability management in the evolving new type power systems With the increasing penetration of renewable energy sources, the system’s fault tolerance has diminished, leading to more intricate transient stability dynamics, especially in hybrid configurations that integrate conventional and renewable generation units. Traditional methods for modeling power system dynamics struggle with modern grid complexities and require substantial computational resources for real-time analysis. This creates a major challenge for fast emergency control responses. To address these issues, this study proposes a dual-driven approach that integrates data-driven insights with model-based strategies for transient power angle control, striking a balance between operational efficiency and economic feasibility. The control framework focuses on minimizing generator tripping and load shedding while ensuring system stability is swiftly restored after severe disturbances. The methodology adopts a deep reinforcement learning paradigm, embedding neural networks with physics-based principles to capture high-order differential interactions between system states and control measures. By forecasting power angle trajectories, this approach mitigates the shortcomings of purely data-driven methods. The proposed decision-making framework not only significantly enhances the system’s ability to recover stability but also improves overall supply reliability. Simulations performed on an modified IEEE-39 bus system showcase the method’s capability and reliability in handling transient stability across systems with differing levels of complexity.

Scenario generation proves to be an effective approach for addressing uncertainties in stochastic programming for power systems with integrated renewable resources. In recent years, numerous studies have explored the application of deep generative models to scenario generation. Considering the challenge of characterizing renewable resource uncertainty, in this paper, we propose a novel two-stage generative architecture for renewable scenario generation using diffusion models. Specifically, in the first stage the temporal features of the renewable energy output are learned and encoded into the hidden space by means of a representational model with an encoder–decoder structure, which provides the inductive bias of the scenario for generation. In the second stage, the real distribution of vectors in the hidden space is learned based on the conditional diffusion model, and the generated scenario is obtained through decoder mapping. The case study demonstrates the effectiveness of this architecture in generating high-quality renewable scenarios. In comparison to advanced deep generative models, the proposed method exhibits superior performance in a comprehensive evaluation.