Is your visual judgment getting closer to the human eye?
In visual inspection sites, have you ever hesitated over whether something is OK or NG?
This time, NANO has added AI functionality to an image sorting machine manufactured by CCM in Taiwan.
(This is the second initiative, following the Unitecho machines from China.)
Until now, we have often heard the following feedback from visual inspection sites.
・Dimensional inspection is fine, but visual pass/fail judgment is unstable.
・Tightening conditions too much leads to excessive false detections.
・Many workpieces have unclear boundaries between OK and NG.
Conventional image processing, in simple terms, makes only binary judgments: OK or NG.
However, real products certainly have a “gray zone” that lies in between.
That is where AI-based learning visual inspection becomes the key.
Using AI is surprisingly simple.
Simply prepare around 20–50 OK images and 50–100 NG images, mark the defect areas, and let the AI learn.
Training time per model is only about 20–30 minutes, making it practical for on-site use.
The AI function can be added simply as an add-on to existing inspection programs.
Models can be switched for each camera, and AI judgment can be turned ON/OFF with a single touch.
As a result, detection reproducibility and stability have greatly improved, especially for small dents and hard-to-quantify appearance defects.
Judgments close to human perception are now becoming possible with machines.
Of course, when it comes to extremely fine scratches or color unevenness,
optimizing lighting conditions and imaging environments remains an important factor.
Even so, there is no doubt that the options for appearance inspection are steadily expanding.
If you are interested in detailed demos or real-world case studies, please feel free to contact us.
Wishing you a great week ahead!
Contact us for comparison and inquiries!
FORTUNE CROSS (THAILAND) CO.,LTD / NANO SEIMITSU CO.,LTD.
Mobile: +66 (0)61-572-1700 Contact : Mr. Yada
#NANO_SEIMITSU ＃AOI_sorting, #CCM_AI
