An industrial visual intelligent quality inspection method for hardware manufacturing
By using a multi-light image acquisition and adaptive feature perception fusion network model, the problems of metal reflection interference and multi-scale defect detection in the quality inspection of hardware products are solved, realizing high-precision and real-time quality inspection and meeting the high-efficiency inspection needs of the production line.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-18
- Publication Date
- 2026-06-16
AI Technical Summary
Existing industrial visual inspection technologies for hardware manufacturing suffer from problems such as interference from metal reflections, difficulty in detecting multi-scale defects, and insufficient inspection speed and accuracy, making it difficult to meet the demands of efficient and high-precision production.
By employing a multi-light image acquisition system combined with a variable-scale adaptive feature perception and fusion network model, a synthetic input image is generated through the registration and dynamic weighted fusion of coaxial light and low-angle light images. A deep learning model with a parallel multi-scale feature extraction structure, a dynamic weight fusion module, and a dual attention mechanism is constructed to achieve high-precision, real-time defect detection of hardware products.
It effectively overcomes interference from metal reflection, adaptively detects defects at multiple scales, improves detection accuracy and generalization ability, meets the real-time requirements of the production line, and achieves efficient quality judgment.
Smart Images
Figure CN122222977A_ABST