Image Sensor Test Device Using Deep Learning for Defect Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for sorting defective image sensors using machine learning and artificial neural networks face challenges in accurately detecting defects due to noise and shading, and require separate test algorithms for different types of image sensors.
Innovation Solution
A test device utilizing a processor to generate a test image through preprocessing operations on raw images from image sensors, and employing a deep learning neural network to classify image patterns and determine whether the image sensor is defective, using a data set that includes normal and defect patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If machine learning and artificial neural networks are applied to the EDS process, then automation and productivity are improved, but measurement precision deteriorates due to noise and shading affecting defect detection accuracy
Solution Approach 1:
The patent segments the defect detection task into multiple specialized neural networks, each trained to detect specific types of defects (e.g., dark blemishes, spot defects, diagonal defects). This segmentation allows each network to focus on particular defect patterns, improving overall detection precision while maintaining automation.
Solution Approach 2:
The patent introduces an intermediary processing stage where raw sensor data is pre-processed and filtered before being fed into the neural networks. This intermediary step reduces the impact of noise and shading by preparing cleaner input data, thereby improving measurement precision without reducing automation.
2Measurement precision
If separate test algorithms are implemented for different types of image sensors, then measurement precision is improved for specific sensor types, but device complexity increases
Solution Approach 1:
The patent develops a universal neural network architecture that can be applied across different types of image sensors. The system uses a common framework with configurable parameters that can be adjusted for different sensor types, maintaining high detection precision while avoiding the complexity of completely separate algorithms for each sensor type.
Solution Approach 2:
The patent employs parameter-based adaptation where the same neural network structure uses different configurable parameters (such as threshold values, filtering coefficients, and training data characteristics) to optimize performance for different sensor types. This approach maintains algorithmic simplicity while achieving type-specific precision through parameter adjustment rather than structural complexity.
Data Source
AI summary
A system for testing whether or not an image sensor included in an image frame is defective includes including a processor configured to execute machine-readable instructions that, when executed by the processor, cause the test device to generate a test image by performing one or more preprocessing operations on a raw image output from an image sensor, and to classify an image pattern of the test image as any one of patterns included in a first data set by using a first deep learning neural network trained based on the first data set and to determine, based on a classifying result, whether or not the image sensor is defective. The first data set comprises a data set for each pattern, classified to correspond to each of one or more defect patterns and to a normal pattern of an image.