Pattern Position Detection Using Integrated Correlation for Unclear Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for positioning components in industrial machines, such as semiconductor chip alignment, are hindered by the need for clear images, which can be obstructed by lighting conditions, leading to reduced accuracy and increased processing time, especially when images are unclear or contaminated.
Innovation Solution
A pattern position detecting method that divides a model image into reference images, aligns and compares these with a detected image using correlation values, integrates these values with spatial filtering to determine positional deviations, and calculates weights based on specific patterns to enhance accuracy and speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image-based positioning is used in clear lighting conditions, then positioning accuracy is high, but processing time increases and the system fails under poor lighting conditions
Solution Approach 1:
The patent divides the image processing into multiple stages: initial rough positioning using correlation methods, followed by refined positioning using feature extraction. The image data is also segmented into different processing streams (correlation-based and feature-based) that operate in parallel, reducing overall processing time while maintaining accuracy.
Solution Approach 2:
The system performs preliminary rough positioning using correlation methods before executing the more time-consuming feature extraction and refinement processes. This preliminary action establishes a starting point that reduces the search space for subsequent detailed analysis, thereby reducing total processing time.
2Ease of manufacture
If traditional target-based positioning is used, then positioning can be performed with simple equipment, but the system requires clear images and stops when images are unclear or targets are missing
Solution Approach 1:
The patent creates a universal positioning system that can handle multiple types of targets and conditions using the same equipment. The system processes both traditional artificial targets and natural feature points through unified correlation and feature extraction algorithms, eliminating the need for separate systems for different target types and improving reliability under varying conditions.
Solution Approach 2:
The system dynamically adjusts processing parameters based on image quality and target detection results. When images are clear, it uses standard processing; when images are unclear or targets are missing, it switches to alternative processing modes with adjusted parameters, maintaining reliability without requiring different equipment.
3Productivity
If image division method is used to reduce processing time, then processing speed increases, but the system requires many learned samples and complex processing steps
Solution Approach 1:
The patent applies partial image division, processing only certain regions or features of the image in detail while using correlation methods for the entire image. This selective approach achieves speed improvements without requiring complete division of all image data, reducing overall processing complexity.
Data Source
AI summary
A pattern position detecting method capable of reducing time for detecting a component position includes: acquiring a model image of a target; dividing the acquired model image into reference images each including a specific pattern; acquiring a detected image of the target; matching origins of the reference images respectively with predetermined positions on the detected image; comparing a region within the detected image with corresponding one of the reference images while moving the origin of the reference image in X and Y directions from the corresponding predetermined position and sequentially acquiring correlation values; integrating the correlation values at respective comparison positions within an integrated XY plane to generate integrated correlation values; and recognizing a value of integrated XY coordinates at a peak of the integrated correlation values as deviation of the specific patterns in the reference images from the predetermined positions of the target within the XY plane.