A hyperbolic aluminum veneer surface flaw detection method based on machine vision
By extracting the aluminum panel region using adaptive gamma transform and U-Net model, and combining median filtering and threshold segmentation, the problem of low contrast in the detection of contaminants on the surface of hyperbolic aluminum panels was solved, achieving higher detection accuracy and stability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUANGDONG YINGJIWEI ALUMINUM BUILDING MATERIALS CO LTD
- Filing Date
- 2026-04-24
- Publication Date
- 2026-07-17
AI Technical Summary
In the detection of contaminants on the surface of hyperbolic aluminum panels, existing technologies suffer from low image contrast due to the similarity between the gray values of contaminants and the background, which affects the accuracy of detection. Furthermore, the gamma transform cannot meet the contrast requirements of different areas, resulting in inaccurate detection.
An adaptive gamma transform method is adopted. By calculating the color feature factor and suspected contamination factor of each pixel, the gamma value is adjusted to enhance the contrast of the contaminant area. The aluminum single board area is extracted by the U-Net semantic segmentation model and median filtering is performed for noise reduction. The contaminant is detected by combining threshold segmentation.
It improves the accuracy and stability of pollutant detection, significantly enhances the contrast of pollutant areas, and makes the detection results clearer and more reliable.
Smart Images
Figure CN122415545A_ABST