Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Hai Wang has a total of 88 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    BMF-DETR: Pseudo-Depth-Guided Bidirectional Multi-Strategy Fusion for End-to-End Object Detection

    by Hai Wang, Junhao Wen, Chunlai Yang, Kamara Kekele Adnan Fayçal, Jiale Gu

    Transformer-based detectors model long-range context effectively, yet their representations remain dominated by RGB appearance and can become unreliable in cluttered, occluded, or crowded scenes. We present BMF-DETR, a pseudo-depth-guided detector that introduces RGB-derived geometric structure without requiring a depth sensor. DA3Mono-Large from Depth Anything 3 generates spatially aligned pseudo-depth maps offline, while two ResNet-50 streams encode appearance and relative geometry. Bidirectional cross-modal attention (BCMA) establishes two-way correspondence, and multi-strategy fusion (MSF) combines the streams through global calibration, channel allocation, and spatial gating before squeeze-and-excitation (SE) recalibration. On the fixed validation/evaluation split of the 2024 Roboflow-curated PASCAL VOC derivative, the complete model reaches 60.80 AP, compared with 52.80 AP for a capacity-matched dual-RGB control. BMF-DETR obtains 49.30 AP on COCO 2017. Its detector contains 58 M parameters and requires 103 GFLOPs; these figures exclude offline pseudo-depth generation. A shared-low-level variant retains 60.10 AP with 50 M parameters and 87 GFLOPs. The results show that pseudo-depth can serve as a useful auxiliary representation when its contribution is separated from capacity effects and evaluated under controlled fusion settings.

  • Open Access

    CSP-UNet: A Lightweight Network for Hand X-Ray Image Segmentation

    by Hai Wang, Jiale Gu, Junhao Wen, Chunlai Yang

    Hand X-ray image segmentation is an important step in automated radiographic image analysis. However, conventional U-shaped segmentation networks often have relatively high model complexity, while variations in grayscale distributions across hand X-ray images may affect segmentation performance. To address these issues, this study proposes a lightweight hand X-ray image segmentation network, CSP-UNet (Cross-Stage Partial U-Net). The network integrates cross-stage partial feature processing into the U-Net encoder–decoder framework to reduce redundant feature computation and the number of model parameters while preserving effective feature representation. In addition, an adaptive Gaussian histogram-matching strategy is employed to reduce variations in grayscale distributions across X-ray images. CSP-UNet was evaluated on a dataset comprising 2000 hand X-ray images and compared with Classic U-Net, Res-UNet, Attention U-Net, and Swin U-Net. Experimental results show that CSP-UNet maintained comparable segmentation performance, achieving a Dice coefficient of 0.9927, PA of 0.9839, MPA of 0.9810, and mIoU of 0.9542, while requiring only 20.01 M parameters. Compared with Classic U-Net, CSP-UNet maintained a comparable Dice coefficient (0.9927 vs. 0.9910) while reducing the parameter count from 69.1 M to 20.01 M, corresponding to a reduction of approximately 71.04%. These results indicate that CSP-UNet maintains comparable segmentation performance while substantially reducing model complexity, offering a favorable trade-off between segmentation performance and model size for hand X-ray image segmentation.

  • Open Access

    Parametric Optimization of a Star-Shaped Bluff Body for Enhanced VIV-Galloping Coupled Energy Harvesting

    by Li Zhang, Hai Wang, Chunlai Yang, Weiwei Duan, Jingjing Peng

    Under low wind speed conditions, conventional bluff body energy harvesters suffer from a single vibration mechanism and a narrow effective wind speed range, making it difficult to meet the continuous power supply demands of miniature electronic devices. In this paper, by systematically optimizing the number of triangular prisms N and the circumferential installation angle α, a parametrically adjustable star-shaped energy harvester (SEH) is proposed. The proposed structure consists of a cylindrical base with a tunable number of triangular prisms uniformly distributed along its circumference, aiming to reveal the regulation mechanism of the VIV-galloping coupling response and energy harvesting performance. Conceptual design and theoretical modeling of the SEH are first carried out. Then, three-dimensional fluid–structure interaction simulations are performed by varying N and α, and a prototype is fabricated for wind tunnel experimental validation. The results show that under the optimal parameter combination of N = 7 and α = 51.4°, the SEH achieves a maximum output voltage of 12.2 V at a wind speed of 3.41 m/s, with a maximum output power of 1.488 mW, and the effective wind speed range is broadened to 2.5~12.44 m/s. Compared with the conventional cylindrical energy harvester (CEH), the SEH (N = 7) increases the maximum output voltage by 44.38%, the maximum output power by 108.4%, and expands the effective wind speed range by 198.50%. Through systematic optimization of key geometric parameters, this study achieves synergistic regulation of flow-induced vibration modes and performance enhancement, providing a parametric design basis for efficient low-speed wind energy harvesting, which can promote the development of self-powered technologies for micro-sensors and IoT devices.


  • A Fluid‐Induced Vibration Energy Harvester Based on Metasurfaces

    by Chunlai Yang, Wenbo Zhu, Hai Wang, Hang Sun, Yimin Lu

    Fluid‐induced vibration (FIV) energy harvesting captures kinetic energy from flowing fluids, converting it into electrical power and showing particular promise for low‐speed wind applications. This study presents a metasurface‐enhanced FIV harvester combined with a branching beam structure to improve energy conversion efficiency. The branching beam adjusts the system's natural frequency, while metasurfaces with circular, square, and striped dimple arrays modify the bluff body's hydrodynamic behavior. Wind tunnel experiments used cylindrical and square cross‐section bluff bodies at flow velocities of 1–6 m s −1 . Results show the circular‐dimple metasurface significantly enhances performance: above 3 m s −1 , it yields output voltages 32%–70.56% higher than conventional harvesters, reaching 1.491 V at 4 m s −1 versus 0.875 V for a smooth cylinder. Maximum output power reaches 2.22 μW, a 188% increase. The striped metasurface improves voltage output by 10%–15%. For square cylinders, the circular‐pit metasurface achieves 4.88 μW at 5 m s −1 , nearly double the conventional output. The branching beam effectively tunes resonant frequency while maintaining piezoelectric efficiency. These findings validate the synergistic use of metasurfaces and structural frequency control for advancing FIV energy harvesting in low‐velocity flows.


  • HyPoseNet: a hybrid CNN for 6D pose estimation with point feature convolution and attention fusion

    by Aolong Jin, Hai Wang, Chunlai Yang, Junhao Wen, Otoide Iretiose Belief, Luyando Stanley Kabalata, Jinsong Gui

    The application of RGB-D image data in tasks such as intelligent perception and pose estimation for robots has recently garnered significant attention. A major challenge is effectively utilizing the complementary modalities in RGB-D data sources. A 6D pose estimation method based on the multimodal fusion method, with a hybrid convolutional neural network (CNN) architecture integrating point feature convolution and attention mechanism is proposed in this study. The method integrated a two-stage framework for segmentation and pose regression of RGB-D data, effectively extracting target features and enhancing the model’s interpretability and robustness. A convolutional block attention module is introduced to recalibrate the feature maps dynamically, which focuses on information-rich regions and channels, thereby improving the overall performance. A series of comprehensive evaluations were carried out on standard benchmark datasets to validate the effectiveness of the proposed method. On the Occlusion LineMOD dataset, it achieved an average ADD accuracy of 96.9%. For the YCB-Video dataset, the method reached a 94.6% average accuracy under the ADD-S AUC metric, and 97.7% of predictions yielded ADD-S distances below 2 cm. These experimental outcomes clearly illustrate the robustness and high precision of the proposed approach.


  • Thermal Analysis of Cylindrical Permanent Magnet Governor with Demagnetization Effect

    by Aolong Jin, Yonglong Zhu, Hai Wang, Henian Li, Chunlai Yang, Jingsong Gui, Ronghua Fu

    Severe heating caused by long-term operation of the cylindrical permanent magnet governor (CPMG) may lead to failure of equipment. The electromagnetic-thermal coupling simulation of a 350 kW CPMG are carried out, which include bi-directional and unidirectional coupling model. The demagnetization of permanent magnets in the CPMG during the temperature rise is considered in the numerical model. The influence of effective coupling ratio and slip rate on the thermal heating of CPMG are studied. The temperature rise during the operation of the CPMG is measured through an experimental platform, which is consistent with the simulation results. Experimental results show that bi-directional coupling model is more reliable compared with unidirectional coupling mode, which can keep the errors within 5%.

  • S2HM: Self-Powered Spine Health Monitoring Using Piezo/Triboelectric Nanogenerators

    by Meng Li, Yongyue Huang, Yanpeng Kan, Jingjing Peng, Sizhong Miao, Zemiao Fang, Xiangzhi Liu, Hai Wang, Xueqin Hu, Min Pan, et al.

    Individuals in professions such as dentistry, aircraft maintenance, and frequent computer usage, are susceptible to developing degenerative changes in the cervical spine due to poor neck posture over time. This can eventually progress to cervical spondylosis. Continuous monitoring of neck health is essential to prevent permanent damage. This study proposes a self-powered, flexible neck brace integrated with four piezoelectric/triboelectric nanogenerators (P/TENGs) designed for the purpose of monitoring neck strength. When the neck undergoes movement, the resultant deformation of the neck brace under force stimulates the P/TENG array, generating voltage output signals from four PENGs and four TENGs. These eight signals are then converted into a 2D intensity map, which is subsequently leveraged for training and prediction through a Convolutional Neural Network (CNN). This approach enables precise differentiation of six distinct neck movements with a precision of 97.78%. The neck brace is integrated and equipped with Inertial Measurement Unit (IMU) sensors to capture neck movement angles and velocities. Combined with data from the P/TENGs, the system offers a comprehensive set of multidimensional data for the evaluation of neck and spine health. Clinical experiments used Principal Component Analysis (PCA) to streamline multidimensional data and applied the KNN algorithm to forecast and categorize cervical curvature abnormality levels (L1-L4), achieving 92.5% accuracy in a trial with 67 participants. In summary, the proposed P/TENG-based neck brace device displays substantial potential for motion recognition and curvature anomaly diagnosis, thereby introducing new prospects for clinical adjunctive diagnosis and home health monitoring.