Detecting Contiguous Defect Regions From Multi-View Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manual inspection of products for quality assurance is time-consuming and prone to accuracy issues, and existing computer vision systems struggle to accurately detect contiguous defect regions across multiple surface views and images.
Innovation Solution
A computer program product and method that processes images of a physical object from different perspectives to detect defect regions, determine their categories, and identify whether they form contiguous regions. The system uses a segmentation model and 3D modeling to calculate accurate spatial metrics of defect regions, accounting for overlapping areas and providing information to a quality assurance module.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual inspection is used to detect defect regions, then accuracy can be maintained through human judgment, but the process becomes time-consuming and inefficient
Solution Approach 1:
The patent replaces manual mechanical inspection with an automated computer vision system that uses machine learning models to detect and analyze defect regions. The system processes images automatically to identify defects, eliminating the need for human inspectors while maintaining or improving detection accuracy through consistent algorithmic application.
Solution Approach 2:
The system creates digital copies of physical objects through imaging to perform defect detection. By working with image representations rather than physical objects, the system enables rapid automated analysis without the time constraints of manual physical inspection, while the machine learning models learn from trained datasets to achieve high accuracy.
2Productivity
If simple defect detection is performed on individual images, then processing speed is maintained, but the system cannot accurately identify contiguous defect regions across multiple views
Solution Approach 1:
The patent transitions from analyzing single 2D images to processing multiple images from different perspectives, adding a dimensional aspect to defect detection. By integrating information across multiple views and perspectives, the system can accurately identify whether defect regions are contiguous or separate, which is impossible when examining individual images in isolation.
Solution Approach 2:
The system merges defect detection results from multiple images and perspectives into a unified analysis. By combining information from different views and using machine learning models to correlate defects across images, the system achieves accurate identification of contiguous defect regions while maintaining efficient automated processing.
3Ease of manufacture
If traditional computer vision methods are used to detect defects, then implementation is straightforward, but the system fails to account for overlapping defect regions and spatial metrics across multiple surfaces
Solution Approach 1:
The patent changes the parameters and metrics used in defect analysis by implementing sophisticated spatial metric calculations that account for defect overlap and three-dimensional surface relationships. The machine learning models are trained to compute accurate spatial metrics including area, volume, and contiguity, moving beyond simple defect presence detection to quantified spatial analysis.
Solution Approach 2:
The system introduces machine learning models as intermediaries between image capture and defect analysis. These models serve as sophisticated processing layers that handle complex spatial relationships, overlapping regions, and multi-surface defect correlations, bridging the gap between simple image input and accurate spatial metric output.
Data Source
AI summary
Provided are a computer program product, system, and method for detecting contiguous defect regions of a physical object from captured images of the physical object. Images are received of a physical object from different perspectives capturing different views of the physical object. Defect regions in the images are detected containing defects on surfaces of the physical object. A determination is made of categories of the defect regions. A determination is made as to whether defect regions of a category have a common boundary to form at least one contiguous defect region for the category. A determination is made of total spatial metric of any contiguous defect regions and non-contiguous defect regions for each of the categories. Information on the total spatial metric for the categories is provided to a quality assurance module to determine a quality of the physical object.