A deep learning-based processing ill-conditioned visual recognition method and system
By extracting the boundary contour and gap information of the defect region in composite materials using a deep learning-based method, and combining the interlaminar shear modulus parameter to calculate the shear stress field, the problem of insufficient accuracy in defect depth discrimination in traditional methods is solved. This enables a comprehensive judgment of defect depth and development trend, thereby improving the quality control capability of composite material processing.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- XIANGTAN INST OF TECH
- Filing Date
- 2026-05-11
- Publication Date
- 2026-07-17
AI Technical Summary
Existing traditional processing pathological visual recognition technology cannot integrate mechanical parameters with visual features in the detection of delamination defects in composite materials, resulting in insufficient accuracy in defect depth discrimination and difficulty in meeting the requirements of high-precision molding quality control.
A deep learning-based approach is adopted. By receiving three frames of images from an industrial camera, a pre-trained deep learning model is used to extract the boundary contours and gap information of the defect area. The shear stress field is calculated by combining the interlayer shear modulus parameter to generate depth discrimination results. Finally, processing ill-condition recognition results are generated by multi-feature fusion.
It achieves stable identification of layered defects, reduces false detections and missed detections, can detect defect expansion trends in advance, enhances the adaptability of the identification scheme and the controllability of the production process, and meets the real-time detection needs of high-speed continuous processing production lines.
Smart Images
Figure CN122415570A_ABST