Display Panel Defect Detection Model Using Ensemble Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current defect detection methods for display panels are inefficient and lack accuracy, particularly in adapting to changing data distributions and varying production lines, which hinders productivity and quality improvement.
Innovation Solution
A detection system comprising a pre-constructed detection model that includes a defect classification identification sub-model and a defect position identification sub-model, utilizing an ensemble learning algorithm and a Convolutional Neural Network model with additional layers to enhance accuracy and adaptability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional defect detection methods are used for display panels, then the detection process is simple, but the detection accuracy and efficiency are low
Solution Approach 1:
The detection model is segmented into multiple specialized sub-models: a defect classification identification sub-model with multiple base models for different defect types, and a defect position identification sub-model. Each sub-model focuses on specific detection tasks, improving overall accuracy while managing complexity through functional division.
Solution Approach 2:
The system dynamically adapts to different production lines and data distributions by training multiple base models with different probability distributions and selectively applying them based on the input data characteristics, enabling the detection system to optimize performance for varying production conditions.
2Adaptability or versatility
If a single detection model is used across different production lines, then the system is simple to maintain, but the adaptability to varying data distributions is poor
Solution Approach 1:
The detection system achieves universality by creating multiple base models that can handle different data distributions and production line characteristics. These models work together through the secondary model to provide a unified detection solution that adapts to various production scenarios without requiring separate systems for each line.
Solution Approach 2:
The system changes parameters by training base models with different probability distributions and sampling ratios. This allows the detection system to adjust its behavior and accuracy based on the specific characteristics of each production line's data, improving adaptability while maintaining a structured multi-model framework.
3Productivity
If manual defect detection is used, then the system complexity is low, but the productivity and detection efficiency are reduced
Solution Approach 1:
The system replaces manual mechanical inspection with an automated detection model that processes images through multiple base models and a secondary classification model. This substitution of mechanical/manual processes with automated computational processes significantly improves detection efficiency and throughput while maintaining high accuracy.
4Measurement precision
If traditional single-model detection is used, then the training process is simple, but the detection accuracy for different defect classifications is insufficient
Solution Approach 1:
The training process is segmented into distinct phases: training multiple base models with different probability distributions and sampling ratios for various defect types, then training a secondary model to integrate their outputs. This segmentation of the training process enables specialized optimization for each defect classification while maintaining overall system coherence.
Data Source
AI summary
The present disclosure provides a detection method. The detection method includes inputting an image to be detected into a detection model being pre-constructed and detecting the image to be detected. The detection model includes a defect classification identification sub-model configured to identify a classification of a defect in the image to be detected, and the defect classification identification sub-model comprises a plurality of base models and a secondary model. The present disclosure further provides an electronic device and a non-transitory computer-readable storage medium.