Visual Inspection Image Sampling for Progressive Library Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated visual inspection systems face challenges in managing massive storage requirements for image data during production lot inspections, as existing sampling techniques either fail to capture a full complement of images across the entire lot or interrupt the inspection process prematurely, leading to incomplete data sets.
Innovation Solution
A progressive update sampling method that ensures a full complement of images is maintained in an image library by replacing images in a manner that distributes them evenly across the inspected lot, allowing for effective fine-tuning and troubleshooting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If only the last X images are saved to reduce storage requirements, then storage space is reduced, but images from the entire lot are not captured and the full complement of samples is not ensured when inspection is interrupted
Solution Approach 1:
The patent changes the sampling parameter from fixed (always saving last X images or one of every Y images) to dynamic (saving every n-th image where n increases progressively). This allows the system to adapt the sampling rate based on the current stage of inspection, ensuring both storage efficiency and comprehensive coverage across the entire production lot.
Solution Approach 2:
The patent implements a dynamic sampling strategy where the sampling interval n is not fixed but changes over time. Initially, n is small to ensure a full complement of images is captured early in the inspection process. As inspection progresses and more images are accumulated, n increases to maintain diversity while managing storage, thereby dynamically balancing completeness and storage requirements.
2Quantity of substance
If one out of every Y images is saved to collect samples from across the entire lot, then storage requirements are reduced, but a full complement of samples is not provided if inspection is interrupted prematurely
Solution Approach 1:
The patent applies preliminary action by initially using a small sampling interval to quickly accumulate a full complement of diverse images at the beginning of the inspection process. This ensures that even if inspection is interrupted early, sufficient training data is available. As inspection continues, the sampling interval increases to maintain storage efficiency while continuing to capture representative samples from across the lot.
3Reliability
If all images are saved to ensure complete data for fine-tuning and troubleshooting, then data completeness is maximized, but storage requirements become massive
Solution Approach 1:
The patent extracts only the necessary subset of images needed for effective fine-tuning and troubleshooting by implementing intelligent sampling. Instead of storing all images, it selectively saves every n-th image where n is determined by the progressive update rule, capturing representative samples across the production lot while discarding redundant images, thereby reducing storage requirements while maintaining data effectiveness.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
A method for sampling images includes receiving a first image set generated by automated imaging equipment during a first inspection period, and storing in a memory an image library that initially consists of the first image set. A plurality of new image sets is then sequentially received during respective inspection periods. While the new image sets are received, the image library stored in the memory is updated. Updating the image library includes, for each new image set, adding to the image library a certain number of images distributed among the new image set and removing from the image library the same number of images distributed among a current instance of the image library. The number of overwritten images in the image library decreases from one inspection period to the next.