AOI Image Classification Using AI Kernel Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated optical inspection (AOI) systems face issues with leakage and overkill, where defects are misclassified, requiring manual intervention and hindering fast production changes due to difficulties in achieving accurate classification.
Innovation Solution
An AI training module is integrated into the AOI system to perform discrete output calculations, apply kernel functions for similarity distance measurement, and weighting analysis to improve classification accuracy, enabling automated determination of sample classification results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If traditional AOI systems use machine visual technology for inspection, then inspection speed is improved, but classification accuracy deteriorates due to leakage and overkill errors
Solution Approach 1:
The patent replaces traditional machine visual inspection algorithms with an AI-based classification system that uses neural networks and kernel functions to process inspection data. This substitution enables the system to maintain high inspection speed while significantly improving classification accuracy by learning from training data and performing sophisticated pattern recognition that traditional algorithms cannot achieve.
Solution Approach 2:
The patent implements a feedback mechanism where inspection results are fed back into the AI training module to continuously improve the classification model. The system uses training characteristic models to learn from past inspection data and refine its classification capabilities, thereby reducing leakage and overkill errors while maintaining high inspection throughput.
2Measurement precision
If traditional AOI systems rely on manual inspection to improve accuracy, then classification accuracy is improved, but productivity deteriorates due to slower inspection speed
Solution Approach 1:
The patent substitutes manual inspection with an automated AI-based classification system that processes inspection data using neural networks and kernel functions. This automation maintains high classification accuracy by using sophisticated algorithms that can distinguish between different defect types and false positives, while simultaneously achieving high production speed through rapid automated processing.
Solution Approach 2:
The patent changes the operational parameters of the inspection system by introducing AI training models that can process and classify inspection data much faster than manual inspection. The system uses training characteristic models that have been optimized to achieve high accuracy in classification while maintaining the speed required for high-volume production.
3Device complexity
If traditional AOI systems use fixed inspection algorithms, then device complexity is reduced, but adaptability deteriorates due to difficulty in achieving fast production change
Solution Approach 1:
The patent introduces dynamic adaptability into the inspection system through AI training models that can be updated and retrained based on changing production requirements. The system uses training characteristic models that can be adjusted to accommodate different product types and defect patterns, enabling fast production changes without requiring complete system redesign or complex reconfiguration.
Solution Approach 2:
The patent implements preliminary training of AI models using training characteristic models before actual inspection begins. This preliminary action allows the system to be pre-configured for specific inspection tasks, enabling rapid adaptation to different production changes without requiring complex real-time reconfiguration of the entire inspection system.
Data Source
AI summary
An automated optical inspection (AOI) image classification method includes sending a plurality of NG information of a plurality of samples from an AOI device into an Artificial Intelligence (AI) module; performing discrete output calculation on the NG information of the samples by the AI module to obtain a plurality of classification information of the samples; performing kernel function calculation on the classification information of the samples by the AI module to calculate respective similarity distances of the samples and performing weighting analysis; based on weighting analysis results of the samples, judging classification results of the samples; and based on the classification results of the samples, performing classification of the samples.