Aperture Ratio Prediction for Display Pixel Deterioration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing display devices face challenges in measuring the aperture ratio of pixels efficiently, leading to luminance variations and afterimages due to uneven deterioration of light emitting elements, which requires costly and time-consuming methods for compensation.
Innovation Solution
An aperture ratio measurement device using machine learning models, such as extreme gradient boost, categorical boosting, and light gradient boost, to predict the aperture ratio of pixels by imaging selected mother substrates, calculating actual and prediction values, and applying ensemble learning for accurate prediction, while also detecting outliers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If actual aperture ratio measurement is performed on all mother substrates, then measurement precision is improved, but productivity is reduced and cost increases
Solution Approach 1:
The patent divides the measurement process into two segments: actual measurement of a small sample subset and machine learning-based prediction for the remaining substrates. This segmentation allows maintaining high measurement precision for the sampled group while achieving high productivity through rapid prediction for the entire batch, resolving the contradiction between precision and speed.
Solution Approach 2:
The patent creates a virtual copy of the measurement process through machine learning models that are trained on actual measurement data. Once trained, these models can rapidly predict aperture ratios for all mother substrates without requiring physical measurement of each one, thus maintaining precision while dramatically improving productivity.
2Measurement precision
If actual aperture ratio measurement is performed on all mother substrates, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent performs preliminary action by measuring a representative sample of mother substrates and using this data to train machine learning models in advance. Once trained, the models can rapidly predict aperture ratios for all substrates without requiring time-consuming actual measurements of each individual substrate, thus maintaining precision while minimizing time loss.
3Productivity
If machine learning prediction is used for all mother substrates, then productivity is improved, but measurement precision may deteriorate
Solution Approach 1:
The patent implements feedback by using actual measurement results from a sample group to train and validate the machine learning models. The models are continuously refined based on the comparison between predicted and actual values, ensuring that prediction accuracy is maintained while achieving high productivity across all mother substrates.
Data Source
AI summary
According to an embodiment of the disclosure, an aperture ratio measurement device includes an actual aperture ratio measurement unit which image some mother substrates selected from a plurality of mother substrates and calculate an aperture ratio actual measurement value of a pixel for the imaged some mother substrates, an estimation aperture ratio measurement unit which learn sensing data for each of the plurality of mother substrates and calculate an aperture ratio prediction value of a pixel for all or a group of the plurality of mother substrates, and an addition unit which calculate a final aperture ratio prediction value based on the aperture ratio actual measurement value and the aperture ratio prediction value.