LED Display Defect Inspection Model Using Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional visual inspection methods for LED displays are labor-intensive and time-consuming, failing to efficiently detect seven common abnormal conditions such as gradient dim line, ghosting effect, and others.
Innovation Solution
A model-building system that preprocesses captured images and uses machine learning algorithms, specifically convolutional neural networks, to build defect inspection models for LED displays, enabling automated detection of abnormalities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If visual inspection is used to check LED display abnormalities, then inspection accuracy can be maintained through human judgment, but labor intensity and time consumption increase significantly
Solution Approach 1:
The patent replaces the manual visual inspection system with an automated image processing system. A camera captures display images, and a processor automatically analyzes them using image processing algorithms to detect the seven types of abnormalities. This substitution eliminates human labor while maintaining detection accuracy through systematic image analysis.
Solution Approach 2:
The system enables the LED display to inspect itself by capturing its own display output and automatically analyzing it for defects. The processor compares the captured images against known defect patterns and provides automated feedback, allowing the display system to perform self-diagnosis without external human intervention.
2Productivity
If automated image processing is implemented for defect inspection, then productivity and speed improve, but system complexity increases due to processing requirements
Solution Approach 1:
The patent segments the defect detection task into seven specific abnormality types (gradient dim line, ghosting effect, non-uniform grayscale level, color shift, LED dead pixel, dim line at first scan line, and high contrast interference). The image processing system analyzes each type separately using targeted detection algorithms, which simplifies the overall complexity compared to attempting to detect all possible defects simultaneously.
Solution Approach 2:
The system uses captured images as copies of the actual display output for analysis. Instead of directly manipulating or testing the physical display components, the system creates digital replicas through camera capture and processes these copies to identify defects, reducing physical system complexity while maintaining detection capability.
3Measurement precision
If comprehensive defect detection covering all seven abnormal conditions is implemented, then measurement precision improves, but device complexity increases due to multiple detection requirements
Solution Approach 1:
The patent implements a universal image processing system that can detect all seven types of abnormalities using a single integrated processor and camera setup. The system processes display images to simultaneously identify gradient dim lines, ghosting effects, grayscale issues, color shifts, dead pixels, scan line defects, and contrast interference, eliminating the need for separate detection devices for each defect type.
Data Source
AI summary
A method of building a model of defect inspection for a light-emitting diode (LED) display is adapted to be implemented by a model-building system. The model-building system stores captured images respectively of LED displays that were displaying images. Each of the captured images corresponds to a status tag that indicates a status of the image being displayed by the respective one of the LED displays. The method includes: performing data preprocessing on the captured images to result in pieces of pre-processed data that respectively correspond to the captured images; and building a model of defect inspection by using an algorithm of machine learning based on the pieces of pre-processed data and the status tags.