Acoustic signal-based garment production auxiliary material quality defect detection and analysis method
By constructing a hierarchical feature analysis framework for acoustic signals and a dual-path supervision mechanism, the problem of unexplainable decision-making by deep learning models in the quality inspection of garment accessories was solved, achieving high-precision and interpretable defect identification and cause analysis, and improving the system's credibility and adaptability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUANGDONG JINDING ZHIZAO GARMENT TECH CO LTD
- Filing Date
- 2026-03-30
- Publication Date
- 2026-07-17
AI Technical Summary
Existing deep learning models in the quality inspection of garment accessories suffer from a "black box" problem, where the decision-making process is not traceable and lacks explanation of physical mechanisms. This makes it difficult to diagnose misjudgments and assign responsibility, affecting the credibility and maintainability of the quality inspection system.
By constructing a hierarchical feature analysis framework based on acoustic signals, the modal response map, energy transfer map, and transient scattering map are separated to extract multidimensional acoustic fingerprint vectors. A lightweight semantic anchoring layer and a dual-path supervision mechanism are introduced to generate interpretable recognition results and a visual traceability view, supporting defect cause analysis and process optimization.
It significantly improves the interpretability and transparency of model decision-making, enhances the robustness and maintainability of the system, achieves high-precision identification capabilities and deep mechanism understanding, and reduces deployment threshold and trust costs.
Smart Images
Figure CN122409871A_ABST