基于DeBruijn条纹编码的多尺度特征融合神经网络相位展开方法及系统
By using a multi-scale feature fusion neural network based on De Bruijn stripe coding, the problem of traditional methods failing to balance global and local features in non-uniform stripe processing is solved. This achieves high-precision phase unfolding without the need for additional auxiliary sequences, making it suitable for 3D measurement systems and improving the robustness and real-time performance of measurements.
CN122223249BActive Publication Date: 2026-07-17HUNAN UNIV
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUNAN UNIV
- Filing Date
- 2026-05-21
- Publication Date
- 2026-07-17
Smart Images
Figure CN122223249B_ABST
Abstract
本发明公开了一种基于De Bruijn条纹编码的多尺度特征融合神经网络相位展开方法及系统，首先构建波长‑De Bruijn元素映射表；随后，利用采集的变形条纹及背景参考图获取包裹相位与背景强度，并据此计算非线性余弦分量与相位梯度特征，从而构造多通道物理特征输入张量。接着，将张量输入至具有多并行支路的多尺度特征融合神经网络。在训练阶段，引入基于码元物理波长加权的损失函数进行约束。本发明通过深度学习网络直接实现相位阶次的高精度预测，无需额外的辅助解码序列，有效避免了传统解相算法中复杂的数学原理过程与繁琐的逻辑检索，显著提升了三维测量的效率与对复杂环境的鲁棒性。
Need to check novelty before this filing date? Find Prior Art