Edge Detection for Irregular Surfaces Using Directional Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Precision machine vision inspection systems face challenges in reliably locating edges on irregular or highly textured surfaces, such as those produced by sawing or laser cutting, due to noise in intensity gradient methods, and existing autofocus methods are complex and not suitable for unskilled users, leading to inaccurate Z-height measurements.
Innovation Solution
The implementation of directional filtering sub regions (DFS) and edge-referenced alignment compensation in machine vision inspection systems, which adjust the directional filtering direction to align parallel to the edge feature, allowing for more accurate and repeatable Z-height determinations by minimizing offset amounts and improving edge profile resolution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If intensity gradient method is used for edge location, then edge detection can be performed, but the method produces noisy results for irregular or highly textured surfaces
Solution Approach 1:
The patent divides the image processing into multiple focus stages, capturing images at different Z-heights. By segmenting the focus depth into discrete steps and analyzing the stack of images, the system can reliably determine edge locations even on irregular surfaces where single-focus intensity gradient methods fail.
Solution Approach 2:
The patent transitions from two-dimensional single-focus image analysis to three-dimensional multi-focus image stack analysis. By adding the Z-dimension (focus depth) to the analysis, the system can distinguish true edge features from surface noise and irregularities, significantly improving measurement precision for textured surfaces.
2Measurement precision
If traditional autofocus methods are used, then Z-height measurements can be obtained, but the methods are complex and not suitable for unskilled users
Solution Approach 1:
The patent implements automated algorithms that perform multi-focus image acquisition and analysis without requiring user intervention. The system automatically captures the image stack, processes the data through focus analysis algorithms, and generates Z-height measurements, making the complex process transparent to unskilled users while maintaining high measurement precision.
Solution Approach 2:
The patent changes the operational parameters of the imaging system by automatically adjusting the focus position through multiple discrete steps. This parameter variation across the Z-axis enables accurate Z-height measurements while the automation hides the complexity of parameter management from the user.
3Measurement precision
If directional filtering is applied to improve edge profile resolution, then measurement precision improves, but the process requires alignment compensation that increases complexity
Solution Approach 1:
The patent performs preliminary alignment compensation by analyzing the image stack to identify the orientation and position of edge features before applying directional filtering. This preliminary action establishes the correct filtering direction, ensuring that subsequent edge profile analysis achieves high resolution without requiring complex real-time alignment adjustments during measurement.
Data Source
AI summary
A method is provided for enhancing edge detection for edges of irregular surfaces in a machine vision inspection system. The inspection system comprises an edge feature video tool configured to determine profile data for an edge feature based on a plurality of differently focused images. An edge-referenced alignment compensation is provided related to substantially minimizing a respective offset amount of the edge feature at respective locations along a directional filtering direction used for directionally filtering the plurality of differently focused images prior to determining the profile data for the edge feature. In some embodiments, the plurality of differently focused images may be directionally filtered using a directional filtering sub region (DFS) defined relative to a point corresponding to a PFF basis pixel location in each of the plurality of images, each DFS having a relatively longer dimension along the directional filtering direction.