Adaptive Threshold Template Matching for Semiconductor Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Template matching in semiconductor wafer inspection faces challenges due to large apparent discrepancies between the template and searched images, leading to inaccurate matching and poor correlation values, making it difficult to determine correct matching positions.
Innovation Solution
The use of mutual feature quantities, calculated from both the template and searched images, to determine matching success or failure, with a scoring system where zero distance indicates correct matching and negative scores indicate incorrect matching, allowing for a fixed score acceptance value.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional template matching is used with fixed threshold values, then matching speed is maintained, but matching accuracy deteriorates when large apparent discrepancies exist between template and searched images
Solution Approach 1:
The patent applies dynamics by making the threshold value adaptive rather than fixed. The threshold is dynamically adjusted based on the distribution characteristics of correlation values obtained during template matching. This allows the system to automatically adapt to varying image conditions and discrepancy levels, improving matching accuracy without requiring manual intervention or complex configuration.
Solution Approach 2:
The system performs self-service by automatically determining appropriate threshold values through statistical analysis of correlation value distributions. The matching process itself generates the data needed to set the threshold, eliminating the need for external calibration or user-defined parameters. This self-adjusting mechanism maintains simplicity while improving accuracy across different matching scenarios.
2Measurement precision
If multiple threshold values are used to improve matching accuracy, then matching precision improves, but the complexity of determining correct matching positions increases
Solution Approach 1:
The patent applies partial action by using a simplified statistical approach (mean and standard deviation) rather than complex multi-threshold analysis. Instead of evaluating multiple threshold scenarios, the system uses a single dynamically calculated threshold based on correlation value distribution, achieving sufficient discrimination between correct and incorrect matches without excessive computational complexity.
Solution Approach 2:
The system changes the parameter of threshold value from a fixed constant to a dynamically calculated value based on correlation distribution statistics. This parameter transformation allows the threshold to automatically adapt to different matching conditions, improving precision while maintaining a straightforward determination process based on statistical measures.
3Adaptability or versatility
If fixed score acceptance values are used, then the matching process is simple and fast, but it cannot adapt to varying image conditions and discrepancy levels
Solution Approach 1:
The patent applies preliminary action by calculating the mean and standard deviation of correlation values during the matching process itself, before final threshold determination. This preparatory statistical analysis enables the system to quickly establish an appropriate threshold based on the actual data distribution, adapting to image variations without requiring time-consuming external calibration or iterative adjustment.
Data Source
AI summary
An inspection device that performs pattern matching on a searched image performs matching between a template image of an inspection object and the searched image by using: a feature region extraction process unit that extracts a feature quantity from the template image acquired for learning; a feature quantity extraction process unit that extracts a feature quantity from the searched image acquired for learning; a mutual feature quantity calculation process unit that calculates a mutual feature quantity of the template image and the searched image from the feature quantity extracted from the template image and the feature quantity extracted from the searched image; a learning process unit that calculates, using a plurality of the mutual feature quantities, a discrimination boundary surface that determines matching success or failure; a process unit that calculates a plurality of the mutual feature quantities from an image acquired from the inspection object; and the plurality of mutual feature quantities and the discrimination boundary surface. Thus, an inspection device can be provided that outputs an accurate matching position in template matching even when there is a large apparent image discrepancy between the template and the searched image.