Visual Inspection Boundary Learning With Pseudo Defect Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing visual inspection devices require human intervention to determine pseudo images as non-defective or defective, making them unsuitable for deep learning applications with feature amounts used in a black-box manner, leading to difficulty in improving determination accuracy.
Innovation Solution
A visual inspection device that includes a storage unit to store pseudo images generated through machine learning, using a boundary learning result between non-defective and defective products, and an inspection unit to inspect objects based on this result, enhancing detection accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human intervention is used to determine whether pseudo images are non-defective or defective, then determination can be made, but it becomes difficult to apply deep learning and improves determination accuracy
Solution Approach 1:
The system performs self-learning by automatically generating pseudo images and using them to refine the determination model without human intervention. The determination unit learns from the generated pseudo images to improve its own accuracy, enabling automated deep learning application.
Solution Approach 2:
Pseudo images are generated in advance as training data before actual determination is performed. This preliminary generation of training data enables the determination unit to learn optimal decision boundaries, improving accuracy while maintaining automation.
2Reliability
If pseudo images are generated for visual inspection, then inspection can be performed, but learning man-hours increase and detection accuracy is limited
Solution Approach 1:
Instead of manually creating diverse training images, the system generates pseudo images by copying and transforming existing images through learned transformations. This automated copying process creates diverse training data without manual effort, improving detection accuracy while reducing learning man-hours.
Solution Approach 2:
The system changes parameters of existing images to generate pseudo images, including transformations such as rotation, scaling, and other parameter modifications. This automated parameter variation creates diverse training data efficiently, improving detection accuracy without increasing manual work.
Data Source
AI summary
Provided are a visual inspection device with an improved determination accuracy of inspection and a method of generating a visual inspection discriminator. The visual inspection device includes: a storage unit configured to store, after generation of, based on a defective product image which includes a defect and which is to be determined as a defective product in visual inspection, at least one pseudo image which is close to a determination criterion between defective products and non-defective products in the visual inspection, a boundary learning result obtained by machine learning of a boundary between non-defective products and defective products through use of a pair of the defective product image and the at least one pseudo image or a pair of two pseudo images; and an inspection unit configured to inspect a surface of an object to be inspected based on the boundary learning result.