PCB circuit board welding spot detection method and system based on image identification

By using non-uniform downsampling of local geometric probability density and a multi-coverage discrimination model, the problems of loss of three-dimensional topological information and difficulty in distinguishing sparse features in existing technologies are solved, achieving efficient detection of minute defects and improving detection accuracy and robustness.

CN122289222AInactive Publication Date: 2026-06-26SHENZHEN RUIDINGRONG TECHNOLOGY CO LTD
1 Cites 0 Cited by

Patent Information

Application Number
CN202610423098.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-04-01
Publication Date
2026-06-26
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing technologies project 3D point cloud data into 2D grayscale images as input to neural networks, resulting in the loss of 3D topological structure and geometric deformation information. This makes it difficult to distinguish between sparse feature points and noise. Furthermore, in industrial scenarios where defect samples are imbalanced, it is difficult to construct clear geometric boundaries, which limits the coverage and discrimination accuracy of defect types with few samples.

Method used

Non-uniform downsampling is performed by calculating local geometric probability density to construct a multi-coverage discrimination model, including local feature vectors and hypersphere space feature vectors. An integrated voting strategy is then used to determine the solder joint quality and defect type.

Benefits of technology

It significantly improves the sensitivity to detect minute defects, reduces the false alarm and missed detection rates, and enhances the detection accuracy and robustness in industrial scenarios with imbalanced samples.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a PCB welding spot detection method and system based on image recognition, and relates to the technical field of image processing.The method comprises the steps that firstly, 3D point cloud of PCB welding spots is obtained and preprocessed, and local geometric probability density is calculated; performing non-uniform downsampling by using density inverse proportion weight, preferentially retaining sparse key features such as welding spot pins and the like, removing redundant data of a substrate, outputting a physical feature point cloud set, inputting the physical feature point cloud set into a feature extraction network, eliminating feature dimension strong correlation through a local and global forging mapping technology, constructing an isotropic hypersphere space feature vector, and extracting the feature point cloud set; and finally, based on the local density and the spatial features, a multi-coverage discrimination model containing the defect hyper-ellipsoid and the normal hyper-ellipsoid is constructed, the welding spot quality is comprehensively judged through an integrated voting strategy, the problems of false detection and missing detection caused by sparse feature loss and sample imbalance are effectively solved, and the detection precision of the small defects is remarkably improved.
Need to check novelty before this filing date? Find Prior Art