一种基于Transformer的湿巾表面缺陷实时检测方法
By using a Transformer-based method that combines wet wipe category information and multispectral feature fusion, the problems of high false negative rate and poor generalization ability in traditional wet wipe surface defect detection methods are solved, achieving more efficient and accurate wet wipe surface defect detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HANGZHOU GUOGUANG TOURING COMMODITY
- Filing Date
- 2025-09-24
- Publication Date
- 2026-07-17
AI Technical Summary
Traditional machine vision-based methods for detecting surface defects in wet wipes suffer from high false negative rates and poor generalization ability. In particular, the reflective noise caused by moisture on the surface of wet wipes is highly similar to the actual defect features, and the surface defects of different types of wet wipes exhibit different characteristics.
A Transformer-based approach is adopted to acquire wet wipe category information, visible light images, and near-infrared images. The approach combines category feature priors with multispectral feature fusion and uses a cross-attention mechanism to dynamically fuse visible light and near-infrared features to generate fused features, suppress reflective noise, enhance real defect features, and filter the detection results through a dynamic confidence threshold.
It improves the accuracy and efficiency of surface defect detection for wet wipes, enhances the generalization ability to different types of wet wipes, reduces interference from irrelevant information, and improves the robustness of the detection model.
Smart Images
Figure CN120976205B_ABST