Two-Stage Industrial Defect Recognition for Small-Defect Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current industrial defect recognition methods based on neural network models are ineffective for detecting small-sized defects and lack precision in localization and size estimation, primarily focusing on single-dimensional defect type recognition.
Innovation Solution
A two-stage region extraction approach using a target region detection model followed by a defect detection model, where the target region is first extracted, and then a defect coarse-selection region is identified, enabling enhanced localization and size estimation of small-sized defects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single-stage defect detection model is used, then the device complexity is low, but the detection precision for small-sized defects deteriorates
Solution Approach 1:
The patent divides the defect detection process into two distinct stages: a target region detection model that identifies potential defect areas, and a defect detection model that performs precise localization and classification within those regions. This segmentation allows each model to specialize in specific tasks, improving overall precision for small defects while managing complexity through functional decomposition.
Solution Approach 2:
The patent extracts and focuses computational resources on regions containing potential defects by first identifying target regions, then applying the more complex defect detection model only to those extracted regions rather than the entire image. This extraction approach improves detection precision for small defects while reducing the effective computational burden.
2Device complexity
If a single recognition dimension for defect type is used, then the recognition system is simple, but the recognition completeness deteriorates
Solution Approach 1:
The defect detection model is designed to perform multiple functions simultaneously: localizing defect positions, estimating defect sizes, and classifying defect types. This multi-functional approach ensures comprehensive defect information extraction without requiring separate systems for each function, balancing recognition completeness with system complexity.
3Productivity
If the entire image is processed directly for defect detection, then the processing speed is fast, but the detection probability for small defects deteriorates
Solution Approach 1:
The patent segments the image processing into two phases: first quickly identifying target regions using a simpler model, then performing detailed defect detection only within those regions. This segmentation maintains overall processing efficiency while significantly improving small defect detection probability by concentrating computational resources where defects are most likely to be found.
Solution Approach 2:
The patent performs preliminary identification of target regions before conducting the main defect detection process. This preliminary action filters out background areas and focuses subsequent detailed analysis on regions containing potential defects, thereby improving detection probability for small defects without substantially increasing overall processing time.
Data Source
AI summary
A target region is first extracted from a to-be-recognized image, and then a defect coarse-selection region including an industrial defect is obtained from the target region. Through two times of region extraction, a proportion of a size of the industrial defect to a size of a background is increased, so that a probability of detecting a small-sized industrial defect can be increased. After the defect coarse-selection region including the industrial defect is detected, localization and size estimation are further performed on the industrial defect in the defect coarse-selection region.