Machine Vision Lighting Setup for Joint Image and Model Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In industrial product appearance checking and other machine learning-based judgment systems, manually optimizing lighting and check algorithms is time-consuming and often fails to achieve optimal accuracy due to the need for frequent adjustments to handle individual workpieces and environmental changes.
Innovation Solution
A method that optimizes both lighting and check algorithm parameters simultaneously using a machine learning model, where captured images and label data are used to update parameters, reducing the need for manual adjustments and improving robustness to environmental changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If manual optimization of lighting and check algorithm is performed, then flexibility to handle individual workpieces is improved, but time consumption and complexity increase significantly
Solution Approach 1:
The patent combines lighting parameters and check algorithm parameters into a single unified optimization process. The machine learning model simultaneously optimizes both lighting conditions and algorithm parameters based on captured images and label data, eliminating the need for separate manual adjustments and reducing overall setup time while maintaining adaptability to individual workpieces.
Solution Approach 2:
The system enables self-service optimization through automated machine learning that performs parameter optimization without requiring manual intervention. The model automatically adjusts lighting and algorithm parameters based on input images and label data, reducing dependence on skilled operators and minimizing time consumption while maintaining high adaptability.
2Ease of operation
If lighting is regulated to be easily observed by operator, then ease of operation is improved, but checking accuracy may not be optimal
Solution Approach 1:
The patent changes the optimization criterion from human visual comfort to machine learning performance. The system optimizes lighting parameters based on their impact on checking accuracy as measured by the machine learning model's loss function, rather than based on operator visual comfort. This allows the system to achieve optimal checking accuracy even when lighting conditions are not ideal for human observation.
3Measurement precision
If multiple captured images with multiple lighting parameters are applied to machine learning model, then optimization accuracy is improved, but calculation amount and system load increase
Solution Approach 1:
The patent applies preliminary action by first training the machine learning model with a large number of captured images under various lighting conditions to establish optimal parameter relationships. Once trained, the model can perform rapid optimization with fewer additional images, as the heavy computational work of learning parameter relationships has already been completed during the initial training phase.
4Device complexity
If lighting parameters and check algorithm parameters are optimized separately, then system complexity is reduced, but overall optimization performance deteriorates
Solution Approach 1:
The patent merges the optimization of lighting parameters and check algorithm parameters into a single unified process. The machine learning model simultaneously adjusts both types of parameters based on the loss function calculated from captured images and label data, ensuring that both parameters are optimized together for maximum checking accuracy rather than separately, which would miss important interactions between them.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure relates to a method, device, system and program for setting a lighting condition when an object is checked and a storage medium. The method includes that: the object is lighted by light sources capable of changing lighting parameters, and the object is captured by an image sensor in such lighting parameters to obtain captured images, wherein the object has known label data; and a part of or all of the captured images and the corresponding label data of the object are applied to learning of a machine learning model, and the lighting condition and the check algorithm parameters of the machine learning model is set simultaneously by optimizing both the lighting parameters and the check algorithm parameters, on the basis of a comparison result between an estimation result of the machine learning model and the label data. Therefore, operations are simplified.