Bin si Wei

Bin si Wei


Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
9 May 2024 in Journal of Real-Time Image Processing
Verified
Journal Article
Closed access
PDF file
CJchao ji
chao ji
MGMingjiang Gao
Mingjiang Gao
XDXinsheng Dong
Xinsheng Dong
BWBin si Wei
Bin si Wei
YZYongcan Zhu
Yongcan Zhu
XHXinbo Huang
Xinbo Huang

16 December 2024 in Journal of Real-Time Image Processing
Verified
Journal Article
Closed access
PDF file
CJchao ji
chao ji
BWBin si Wei
Bin si Wei
FZFan Zhang
Fan Zhang
XJXinghai Jia
Xinghai Jia
ZSZhiwei Song
Zhiwei Song
XHXinbo Huang
Xinbo Huang

12 September 2024 in Journal of Real-Time Image Processing
Verified
Journal Article
Closed access
PDF file
XJXinghai Jia
Xinghai Jia
CJchao ji
chao ji
FZFan Zhang
Fan Zhang
BWBin si Wei
Bin si Wei
MGMingjiang Gao
Mingjiang Gao
XHXinbo Huang
Xinbo Huang

30 January 2025 in Research Square
Verified
Preprint
Open access
PDF file
CJchao ji
chao ji
ZLZhenyu Li
Zhenyu Li
BWBin si Wei
Bin si Wei
SZSiyuan Zhou
Siyuan Zhou
JHJiayi He
Jiayi He
XHXinbo Huang
Xinbo Huang
The vibration dampers on transmission lines, after prolonged exposure to natural environments, are prone to corrosion and damage caused by various external factors. Detection algorithms specifically designed for small targets like vibration dampers are still scarce. Moreover, existing algorithms often exhibit drawbacks such as limited accuracy and insufficient real-time performance.To address these challenges, this paper proposes a novel detection method for abnormal vibration dampers on transmission lines based on an EffDetNet framework. Specifically, Multi-dimensional Separable Concatenated Residual Convolution module (MDSCR) is introduced to enhance contextual connections through multi-scale feature extraction. By integrating feature fusion, the model preserves more detailed features, while depthwise separable convolution reduces computational overhead, improving both accuracy and detection speed. Additionally, a Dynamic Allocation Attention mechanism (DAAM) is proposed to effectively capture diverse dimensional information, thereby enhancing feature focus and perceptual capabilities.To further optimize performance, we design a lightweight re-parameterized convolution and shared convolution detection head (LRSC), which significantly reduces model parameters and enhances computational efficiency. Finally, we adopt a Shape-IoU bounding box regression loss function to achieve faster convergence and higher detection precision.Experimental results validate the high efficiency and performance of the EffDetNet model. On the dataset constructed for this study, the model achieves a mean Average Precision (mAP) of 94.3% and a Frames Per Second (FPS) rate of 128.3. Leveraging this model, we developed an abnormal vibration damper monitoring system for transmission lines, demonstrating significant practical engineering value.