Circuit board assembly defect feature recognition method based on deep learning

CN122265748APending Publication Date: 2026-06-23HUNAN AVIONICS XINTONG TECHNOLOGY CO LTD
View PDF 1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
HUNAN AVIONICS XINTONG TECHNOLOGY CO LTD
Filing Date
2026-05-27
Publication Date
2026-06-23

AI Technical Summary

Technical Problem

Existing deep learning models struggle to effectively distinguish between deterministic physical reflections and random defect interference in the optical inspection of highly reflective metal components, resulting in poor stability in identifying minute defects. Furthermore, existing improved methods are prone to introducing identification risks and lacking generalization ability.

Method used

By acquiring image data and spatial topology design data of circuit board components, a prior density tensor for spatial topology alignment is generated. Local standard deviation is calculated using a sliding window to generate a contrast excitation mask, suppressing high-reflection artifact components and compensating for edge feature component responses. A deep neural network is then optimized by combining contrast-aware constraint terms.

Benefits of technology

It achieves stable identification of minute defects under highly reflective backgrounds, enhances the adaptability of detection, reduces the false alarm rate, and improves the identification accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122265748A_ABST
    Figure CN122265748A_ABST
Patent Text Reader

Abstract

The application relates to the technical field of image recognition, and discloses a circuit board assembly defect feature recognition method based on deep learning, which comprises the following steps: acquiring image data of a detection target and extracting a feature tensor; reading spatial topological design data and projecting geometric wiring to a feature tensor coordinate space to generate a spatial topological alignment prior density tensor; applying a nonlinear damping penalty to a local standard deviation according to a numerical distribution to generate a contrast excitation mask; correcting the feature tensor by using the mask, suppressing a high reflection artifact component and enhancing an edge feature, and outputting an enhanced feature tensor to recognize defects; and coupling physical design priori and visual features, suppressing physical artifacts generated by metal reflection, compensating for edge micro signals, and maintaining feature expression constancy under cross-illumination working conditions.
Need to check novelty before this filing date? Find Prior Art