AI Inspection Thresholding for Intermediate Defect Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI-based abnormality detection systems lack flexibility in distinguishing between normal and abnormal objects based on the severity of the abnormality, requiring multiple machine learning data sets for different applications and failing to account for intermediate quality objects.
Innovation Solution
A computer-readable recording medium stores an inspection program that uses a machine learning model to set thresholds for determining abnormal levels, allowing flexible selection of objects as normal, abnormal, or intermediate based on the severity of the abnormality, and adjusts ejection rates for intermediate objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single machine learning model is used for abnormality detection, then the device complexity is reduced, but the adaptability to different applications and abnormality severity levels deteriorates
Solution Approach 1:
The patent applies parameter changes by introducing an abnormal level threshold parameter that can be adjusted according to different applications and requirements. Instead of using multiple machine learning models for different abnormality severity levels, the system uses a single model with configurable threshold parameters (e.g., first threshold, second threshold) to classify objects into normal, intermediate, and abnormal categories. This allows the same model to adapt to different applications by changing the threshold values rather than changing the model itself.
2Device complexity
If traditional binary classification is used, then the classification process is simple, but the ability to identify intermediate quality objects deteriorates
Solution Approach 1:
The patent applies segmentation by dividing the traditional binary classification (normal/abnormal) into three distinct segments: normal objects, intermediate quality objects, and abnormal objects. This is achieved by introducing an intermediate category between normal and abnormal, with specific threshold ranges for each category. The segmentation allows the system to identify and handle intermediate quality objects that were previously misclassified as either normal or abnormal, thereby improving measurement precision without excessive complexity.
3Measurement precision
If multiple thresholds are set for different abnormal types, then the measurement precision of abnormality detection is improved, but the device complexity increases
Solution Approach 1:
The patent applies universality by designing a single threshold configuration system that serves multiple functions: it classifies normal vs. abnormal objects, identifies intermediate quality objects, and handles multiple abnormal types simultaneously. The first threshold and second threshold work together to create three classification zones that can accommodate different abnormal types (e.g., scratches, dents, discolorations) without requiring separate threshold sets for each type. This universal threshold system reduces device complexity while maintaining high measurement precision.
Data Source
AI summary
A non-transitory computer readable recording medium has stored therein an inspection program that causes a computer that is to inspect a target object to execute a method, the method including setting a threshold to specify whether the target object is a normal object or an abnormal object in accordance with an abnormal level of the target object, determining an abnormal level of the target object by applying a captured image of the target object to a machine learning model, and specifying the target object to be the abnormal object if the target object includes a part having the determined abnormal level equal to or more than a threshold set for each of a plurality of abnormal types having possibility of occurring at a same part of the target object.