Synthetic Image Generation for Automated Visual Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated visual inspection systems for pharmaceuticals and other applications face challenges in developing robust image libraries due to the need for large, diverse, and balanced image sets to avoid false negatives and false positives, especially when dealing with small or bland defects, which is labor-intensive and costly, and requires frequent updates with product or process changes.
Innovation Solution
The use of arithmetic transposition algorithms and deep learning-based inpainting techniques to generate synthetic images that are causally representative, adding variability to training libraries, and assessing image suitability through quality control techniques to improve model performance and reduce false rejects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If traditional image transformation techniques (reflection, linear scaling, rotation) are used to expand image libraries, then the image library size increases, but the images lack causal representation and balance, leading to poor model performance
Solution Approach 1:
The patent uses synthetic image generation techniques to create copies of defect images with varied characteristics. Instead of simply transforming existing images, the system generates new synthetic images that replicate defect patterns while introducing controlled variations in background, lighting, and defect positions, thereby expanding the image library with causally representative samples
Solution Approach 2:
The system systematically varies multiple image parameters including defect size, defect position, background characteristics, lighting conditions, and image noise levels. These parameter changes are applied in a controlled manner to generate diverse yet realistic images that maintain causal relationships between defects and their visual representations, improving both library size and model reliability
2Reliability
If large and diverse training image libraries are created, then model coverage improves, but the development process becomes highly iterative, complex, labor-intensive, and costly
Solution Approach 1:
The system implements automated workflows where the synthetic image generation process is self-directed through algorithmic control. The system automatically generates images, applies transformations, and creates augmented datasets without requiring manual intervention for each image, thereby reducing labor intensity and process complexity while maintaining diverse and comprehensive image libraries
Solution Approach 2:
The synthetic image generation system serves multiple functions: it expands image library size, introduces diversity in defect representations, balances class distributions, and provides controlled variation for robust model training. This multi-functional approach consolidates several development tasks into a single unified process, reducing overall complexity
3Adaptability or versatility
If product line changes or inspection process changes occur, then system adaptability is required, but complete image library rebuild is necessary, increasing time and cost
Solution Approach 1:
The system employs dynamic image generation capabilities that can adapt to product line changes and inspection process modifications in real-time. When changes occur, the synthetic image generation process automatically adjusts parameters and generates new relevant images without requiring a complete library rebuild, enabling rapid system adaptation while minimizing time loss
Data Source
AI summary
Various techniques facilitate the development of an image library that can be used to train and/or validate an automated visual inspection (AVI) model, such an AVI neural network for image classification. In one aspect, an arithmetic transposition algorithm is used to generate synthetic images from original images by transposing features (e.g., defects) onto the original images, with pixel-level realism. In other aspects, digital inpainting techniques are used to generate realistic synthetic images from original images. Deep learning-based inpainting techniques may be used to add, remove, and/or modify defects or other depicted features. In still other aspects, quality control techniques are used to assess the suitability of image libraries for training and/or validation of AVI models, and/or to assess whether individual images are suitable for inclusion in such libraries.