Examples
Shape inspection using deep learning
| Category | Electrical components |
|---|---|
| Workpiece | Electrical components |
| Inspection details | Soldering shape inspection |
| Application | Deep Learning |
Before
The weld shape differs every time, so without being able to create a standard for defect detection, it has been impossible to detect defects using rule-based image processing. Thus, we have had to rely on human visual inspections.
After
By applying deep learning, it has become possible to learn what defective shapes are and detect defects as good as the human eye can. Also, it is possible to save the images of and history of inspection results, which is difficult to do with visual inspections.
[Main characteristics]
In image inspection using deep learning, the setting and adjusting of algorithms, filters, and threshold values for conventional image processing require many man-hours, and there is no notion of adjusting them. It is possible to perform high precision inspections by training this system using only OK images, or a mix of OK and not OK images.
*Our company uses VisionPro Vidi by Cognex.
Also, this system can be used for complicated application which tended to be a high-cost, difficult hassle for conventional machine vision systems. Deep learning, which is flexible when dealing with deviation and the variance of defects, boasts functions that exceed conventionally high-level quality inspections, especially in the following applications.
- Defect detection
- Surface treatment/material classification
- Assembly verification/deformed area detection
- Character recognition (including distortions in printing)
In image inspection using deep learning, the setting and adjusting of algorithms, filters, and threshold values for conventional image processing require many man-hours, and there is no notion of adjusting them. It is possible to perform high precision inspections by training this system using only OK images, or a mix of OK and not OK images.
*Our company uses VisionPro Vidi by Cognex.
Also, this system can be used for complicated application which tended to be a high-cost, difficult hassle for conventional machine vision systems. Deep learning, which is flexible when dealing with deviation and the variance of defects, boasts functions that exceed conventionally high-level quality inspections, especially in the following applications.
- Defect detection
- Surface treatment/material classification
- Assembly verification/deformed area detection
- Character recognition (including distortions in printing)