betway必威(中国)官方网站

2D/3D Mesh Defect Inspection

Solution brief

The 2D/3D mesh defect inspection solution based on AI computer vision adopts TimesAI deep learning development platform independently developed by Dihuge, integrating CV+AI+Automation, to achieve online real-time appearance defects detection, defective product removal, yield management, and other aspects of 3C mesh precision small parts.

The system effectively solves three major difficulties in the field: 1. The difference in the occurrence probability of different types of defects is large, and imbalance of defect samples. 2. Defects are small target defects. 3. Defect types have different semantic levels.

Currently, the system covers dozens of defect types, such as: steel mesh hull deformation, bending deformation, foreign object, clamping waste, burr, hull mesh breaking, wire mesh separation, steel mesh offset position, steel mesh deformation, steel mesh slag, wavy edges, large and small edges, liner offset position, liner more product, leaky white, peeling glue, lack of glue, overflow glue, glue offset position, and so on.

Solution function

The 2D/3D Mesh defect inspection based on AI computer vision is a one-stop solution for 3C mesh mounted precision small parts, which mainly has online real-time appearance defect detection, defective product removal, yield management and other functions. The system performance index of this plan is higher than that of the industry, and the plan is constantly iterated and upgraded, which can replace and outperform the manual, improve the yield, reduce the staff and increase the efficiency.

Bright spot

This solution effectively solves the three major difficulties in defect detection (unbalanced defect samples, small target defects, different semantic levels of defect types).

Solve the problem of unbalanced defect samples
Solve the problem of small target defect detection
Solve the problem of different semantic levels of defect types
Excellent performance
One-stop service
Reduce personnel and increase efficiency

Application scenario

2D/3D mesh precision small parts cosmetic defect inspection, sorting, and assembly

Product case

The Mesh Precision Parts Detection/Packaging Integrated Machine developed by the top three mobile phone companies in the global industry is based on the TimesAI deep learning development platform of Dihuge, which can quickly build a detection system and is appropriate for various line structures and sorting logic of users. It can detect more than 20 type detects, such as metal hull deformation, bending deformation, foreign object, clamping waste, burr, metal impress, metal scratch, hull mesh breaking, wire mesh separation, steel mesh offset position, steel mesh deformation, steel mesh slag, steel mesh damage, wavy edges, large and small edges, liner offset position, liner more product, leaky white, peeling glue, lack of glue, overflow glue, glue offset position. Defect detection performance index is : Escape＜0.5%, overkill＜3%.

Scenario
Scenario
Scenario