Image Inspection Inference Using Pretrained Defect Boundaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI image sensors for workpiece inspection require significant time and effort to establish learning-based inspection modes, necessitating retraining for different production lines, limiting usability.
Innovation Solution
An image inspection apparatus with a built-in illuminator, camera, and learned neural network storage that prepares and stores neural networks for both failure/no-failure determination and classification, allowing for efficient inference without retraining.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a learning-based inspection mode is established using a neural network, then inspection accuracy and adaptability are improved, but the time and effort required for training and establishing the neural network increases significantly
Solution Approach 1:
The patent applies preliminary action by pre-training neural networks for multiple inspection scenarios before actual use. The system prepares multiple neural networks in advance, each trained on different types of defects or inspection conditions, so that when inspection is needed, the pre-trained networks are already available for immediate deployment without requiring time-consuming training at the point of use.
Solution Approach 2:
The patent implements universality by designing a neural network system that can handle multiple inspection tasks with a single unified architecture. The system uses a common neural network structure that can be configured for different inspection types through parameter adjustments rather than requiring completely separate networks, reducing overall training time and effort while maintaining high accuracy across different inspection scenarios.
2Adaptability or versatility
If separate neural networks are trained for different production lines or inspection types, then adaptability to specific conditions is improved, but the complexity of managing and retraining multiple networks increases
Solution Approach 1:
The patent applies universality by creating a unified neural network management system that can handle multiple production lines and inspection types through a single platform. The system uses a common neural network architecture with configurable parameters that can be adjusted for different inspection scenarios, eliminating the need to manage completely separate networks for each production line while maintaining specific adaptability.
Solution Approach 2:
The patent uses parameter changes to adapt the neural network to different production lines and inspection types. Instead of training entirely new networks for each scenario, the system adjusts specific parameters and hyperparameters of the existing neural network architecture to suit different inspection conditions, thereby reducing management complexity while preserving adaptability.
3Measurement precision
If a neural network is retrained for each new inspection scenario, then inspection quality for specific conditions is improved, but productivity and ease of operation deteriorate due to repeated training requirements
Solution Approach 1:
The patent applies preliminary action by pre-training neural networks for various inspection scenarios in advance. Multiple neural networks are trained beforehand on different defect types and inspection conditions, so that when actual inspection is needed, the pre-trained networks are immediately available for deployment without interrupting production for retraining, thus maintaining both high inspection quality and productivity.
Solution Approach 2:
The patent uses copying by creating multiple pre-trained neural network models that can be copied and deployed across different inspection scenarios. Instead of retraining from scratch for each new scenario, the system copies existing pre-trained networks and adapts them through parameter adjustments, thereby maintaining inspection quality while avoiding the productivity loss associated with repeated full retraining cycles.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Facilitates quick and efficient inspection by eliminating the need for retraining neural networks, creating a simple environment for both failure/no-failure determination and classification.
Implementation Method 1
The illuminator irradiates a workpiece as an inspection object with illumination light. The camera receives light that is reflected from the workpiece
Data Source
AI summary
An image inspection apparatus includes a learned neural network storage storing a neural network that previously learns weighting factors between input, intermediate and output layers, and an inferer determining failure/no-failure of a workpiece and classify the workpiece to classes based on an image of the workpiece. The inferer performs first and second inferences. In the first inference, the inferer determines failure/no-failure of the workpiece based on failure/no-failure feature quantities that are obtained by providing the workpiece image to the neural network and a failure/no-failure determination boundary. In the second inference, the inferer define a classification boundary to be used to classify an inspection workpiece to the classes in a feature quantity space of the neural network based on classification feature quantities that represent the different-type classification workpiece images, and classifies a workpiece to the classes based on classification feature quantities of an image of the workpiece and the classification boundary.