Topology Normalization for Anomaly Detection in X-Ray Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image anomaly detection techniques face challenges in accurately identifying defects in objects during non-destructive testing, particularly due to variations in anomaly scores that are not observed in training data, such as those caused by x-ray diffraction and image alignment issues, leading to noisy and unreliable results.
Innovation Solution
The system defines alternative topologies based on anomaly-score data and neighborhoods, normalizes anomaly scores using topology mappings, and employs multi-stage post-processing with adaptive thresholds to improve the detection of anomaly regions and sizes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional anomaly detection techniques are used, then the detection process is simple, but the accuracy is reduced due to noisy anomaly scores from x-ray diffraction and alignment issues
Solution Approach 1:
The patent segments the anomaly detection process into multiple stages: initial anomaly scoring, topology construction based on score neighborhoods, multi-stage post-processing with different thresholding strategies, and statistical outlier removal. This segmentation allows each stage to address specific aspects of the problem, improving overall accuracy while keeping individual steps manageable.
Solution Approach 2:
The patent introduces topology as an intermediary structure that mediates between raw anomaly scores and final defect identification. By constructing topologies based on neighborhoods of anomaly scores and using topology mappings, the system transforms noisy score data into a structured representation that preserves spatial relationships and enables more reliable defect detection.
2Reliability
If adaptive thresholding and multi-stage post-processing are applied, then false positives are reduced, but processing time increases
Solution Approach 1:
The patent performs preliminary actions by constructing topologies and computing neighborhood statistics before final thresholding. The multi-stage post-processing approach prepares adjusted anomaly scores in advance using topology-based normalization, so that when defects need to be identified, the scoring is already optimized, reducing the time penalty of multiple processing stages.
Solution Approach 2:
The patent applies partial processing to different regions of the image based on their anomaly characteristics. Not all regions undergo the full multi-stage post-processing pipeline; instead, processing intensity is adapted to local needs, with some regions receiving simplified treatment. This selective approach maintains reliability for critical defect regions while reducing overall processing time.
Data Source
AI summary
Some embodiments of devices, systems, and methods obtain at least one first image, wherein the at least one first image is defined in an image space; select at least one feature in the at least one first image; define a topology based on the at least one feature; generate a topology mapping between the topology and the image-space topology; obtain a plurality of anomaly scores, wherein each anomaly score of the plurality of anomaly scores was generated based on a respective detection area in a second image; map the plurality of anomaly scores to the topology based on the topology mapping; and normalize each anomaly score in the plurality of anomaly scores based on the respective neighboring anomaly scores in the topology, thereby generating normalized anomaly scores.