Image template matching method for wafer image and related product
By using a machine learning-based algorithm to extract wafer image feature information and automatically selecting the best algorithm combination for image template matching, the problem of low efficiency and unstable accuracy in existing technologies is solved, and efficient and accurate wafer image matching is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- DONGFANG JINGYUAN ELECTRON LTD
- Filing Date
- 2026-03-30
- Publication Date
- 2026-06-19
AI Technical Summary
Existing wafer image template matching methods rely on human experience, resulting in low efficiency and unstable accuracy.
A machine learning-based algorithm is used to select a model to extract feature information from wafer images, and the combination with the highest matching accuracy is selected from multiple candidate algorithm combinations for image template matching.
It improves the efficiency and accuracy of wafer image template matching, reduces human intervention, and the output results are not affected by subjective experience. Furthermore, the accuracy can be continuously improved through training.
Smart Images

Figure 1 
Figure 2