Visual Inspection Pattern Positioning via Segmented Sorting Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing visual inspection methods face challenges in accurately positioning a feature portion of interest from an inspection target image due to numerous similar candidates, which can lead to incorrect detection and sorting issues, especially in applications like robot-picking and alignment marking.
Innovation Solution
A positioning method that involves setting a standard region around the standard pattern and a sorting region in the inspection image to evaluate and sort candidates based on their similarity, using evaluation values to accurately identify and position the desired feature portion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If edge information from the standard image is used to find candidates in the inspection target image, then the search capability is improved, but many similar candidates are extracted including irrelevant ones, reducing positioning accuracy
Solution Approach 1:
The patent divides the standard image into multiple regions including a standard pattern region and a sorting region. By segmenting the search space and introducing a separate sorting region for evaluation, the system can extract multiple candidates efficiently while then evaluating and ranking them to identify the most relevant one, thus resolving the contradiction between finding all similar candidates and accurately positioning the desired feature.
2Reliability
If multiple candidates similar to the standard pattern are extracted from the inspection target image, then the completeness of detection is improved, but the complexity of sorting and evaluating candidates increases
Solution Approach 1:
The patent introduces a sorting region as an intermediary element that serves as a basis for evaluating candidates. This sorting region acts as a mediator between the extracted candidates and the final selection process, providing an objective criterion (similarity to the sorting region) for ranking candidates without requiring complex sorting algorithms, thus reducing the overall system complexity while maintaining detection completeness.
3Measurement precision
If a standard region is set to surround the standard pattern, then the feature extraction is improved, but the ability to distinguish the desired feature from similar candidates deteriorates
Solution Approach 1:
The patent segments the standard image into distinct regions: a standard pattern region for feature extraction and a separate sorting region for candidate discrimination. This segmentation allows the system to maintain high feature extraction accuracy from the standard pattern while simultaneously using the sorting region as a reference to evaluate and distinguish the desired candidate from similar ones, thus resolving the contradiction between precise feature extraction and reliable candidate discrimination.
Data Source
AI summary
A feature portion desired by a user out of an inspection target image is accurately positioned. A standard region is set so as to surround a standard pattern in a standard image of a product to be a standard for an inspection target. A sorting region, which is a region for sorting a plurality of candidates similar to the standard pattern, is set in the inspection image. The standard pattern is searched from the inspection image, to extract a plurality of candidates similar to the standard pattern. The sorting region is disposed with respect to each of the plurality of candidates for the standard pattern, extracted in the extraction step, to sort a candidate for the standard pattern based on an evaluation value of the sorting region disposed with respect to each of the plurality of candidates for the standard pattern.