Circuit board multi-type defect intelligent identification method and system based on deep learning

By combining deep learning with multi-view consistency fusion and distribution difference analysis, the problem of unstable identification results in circuit board defect detection is solved. This method achieves stable differentiation between process-allowed fluctuations and real functional defects, as well as suppression of false defects, thereby improving the stability and accuracy of detection.

CN122156728APending Publication Date: 2026-06-05GUANGZHOU ELECTROMECHANICAL SENIOR TECHN SCHOOL
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
GUANGZHOU ELECTROMECHANICAL SENIOR TECHN SCHOOL
Filing Date
2026-02-05
Publication Date
2026-06-05

AI Technical Summary

Technical Problem

Existing circuit board defect detection technologies struggle to reliably distinguish between process-permissible fluctuations and genuine functional defects in scenarios where regular structures and irregular anomalies are mixed, leading to unstable identification results and limited adaptability to changes in imaging conditions and data distribution drift.

Method used

By using a deep learning-based intelligent identification method for multiple types of defects, printed circuit board inspection data and environmental historical auxiliary data are collected and preprocessed. Process semantic separable density ratio discrimination analysis is performed. Combined with multi-view consistency fusion and distribution difference joint drift alarm analysis, stable differentiation between process allowable fluctuations and real functional defects and closed-loop correction of imaging parameters are achieved.

Benefits of technology

It effectively solves the problem of unstable recognition results under mixed distribution of regular structures and irregular anomalies, suppresses false defects caused by reflection, occlusion and viewing angle differences, realizes effective characterization of small structural anomalies and weak contrast defects, and improves the consistency and reliability of detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122156728A_ABST
    Figure CN122156728A_ABST
Patent Text Reader

Abstract

The application discloses a circuit board multi-type defect intelligent identification method and system based on deep learning, and relates to the technical field of industrial vision.The circuit board multi-type defect intelligent identification method and system based on deep learning comprises the following steps: S1, preprocessing printed circuit board detection data and environmental history auxiliary data; S2, triggering re-sampling consistency verification and entering multi-view consistency fusion verification request according to process semantic separable density ratio discriminant analysis results; S3, performing information consistency and spatial consistency joint checking and multi-view consistent and reliable judgment, and executing consistency disambiguation disposal and review shunting; and S4, executing imaging closed-loop correction and incremental learning triggering according to distribution difference joint drift alarm analysis results.The application solves the problem that the existing circuit board defect detection method is difficult to stably distinguish between process allowable fluctuation and real functional defects in the scene of mixed distribution of regular structure and irregular abnormalities, and is prone to unstable identification results.
Need to check novelty before this filing date? Find Prior Art