Bridge cable disease millimeter-level positioning method and system based on small sample learning
By employing a method combining few-shot learning and deep information fusion, and using a twin feature extraction network and attention mechanism module, the problems of data dependence and low positioning accuracy in bridge cable defect detection are solved, achieving pixel-level precise segmentation and millimeter-level physical size positioning of bridge cable defects.
CN122415508APending Publication Date: 2026-07-17WUHAN HONGHAIXIN TECH CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- WUHAN HONGHAIXIN TECH CO LTD
- Filing Date
- 2026-04-21
- Publication Date
- 2026-07-17
Smart Images
Figure CN122415508A_ABST
Abstract
本发明涉及计算机视觉与结构健康监测交叉技术领域，具体公开了一种基于小样本学习的桥梁拉索病害毫米级定位方法、系统，包括：获取支持集图像、查询集图像及深度信息；采用截断式双生特征提取网络输出保留高分辨率空间信息的高维特征张量；通过注意力机制模块以支持集特征为参考对查询集特征进行空间维度上的特征交互与融合；将联合特征图输入基于卷积神经网络的关系度量网络生成响应图并输出定位信息；结合深度信息及相机内参将像素尺寸转化为绝对物理尺寸。本发明采用小样本学习范式，仅需少量参考样本即可识别未知病害，通过高分辨率特征保留与交叉注意力机制克服拉索曲面导致的视角变形，融合深度信息实现毫米级物理尺寸定位。
Need to check novelty before this filing date? Find Prior Art