OV20i Use Case

The OV20i is an AI powered all-in-one vision system

ACCURATE

Outperforms other traditional and AI vision systems and sensors on accuracy and speed benchmarks.

EASY

No special cameras or external lighting. Software runs directly on the device with no downloads or internet connections.

POWERFUL

Our algorithms identify defects even on shiny, rounded, organic and difficult-to-inspect parts.

RELIABLE

All-in-one hardware based on NVIDIA technology gives the OV20i industry-leading uptime.

AI powered all-in-one vision system

With an onboard GPU, it allows you to create and deploy an automated inspection program in 30 minutes using only 20 samples, without software downloads or programming experience.

Challenge

Transparent or translucent parts can be very difficult to see with vision systems. Even with creative lighting and optics, the classical vision tools themselves often struggle even with simple presence checks. This is compounded when your clear parts aren’t always in the exact same location, like when loading a box with 4 clear glasses and one translucent air cushion.

Solution

Using the photometric mode of the imaging setup, you are able to get more stable images even with the clear glasses. While the images change slightly with ambient lighting changes such as an overhead light turned on or off or a flashlight directed at the box, you can train the AI model to plan for such variation.

By defining regions of interest (ROIs) that cover each interior corner of the box and extend to the center, you allow the glasses to be in slightly different positions and still check for presence. The same goes for the green-tinted air cushion, with an ROI that encompasses most of the places it could squeeze in between the glasses.

By having two groups of ROIs, one for the 4 individual glasses and one for the air cushion, you can gather more training data from far fewer images. The label training images within each grouped ROI allows the deep-learning algorithms to build an AI model using all of the regions shown.

Results

After taking enough images in different conditions with all, some, or none of the glasses missing, you can train a reliable AI model. Labeling training images with the glasses and the air cushion in different positions and with different lighting makes the final AI model even more robust.

This results in judgments that are confident when one or more glasses is missing or if the air cushion is absent.

Conclusion

Without the OV20i’s advanced AI, you would have a tough time seeing the presence of the glasses and accounting for changes in ambient lighting and positioning inside the box for the glasses and air cushion.

This same advantage to the OV20i would apply to checks of other transparent parts, especially when ambient light and variable positioning can be a factor.

Ready to see what Overview can do for you?

Invest in reliable technology to hone your factory’s competitive edge. Schedule a camera and industrial vision software demo today. Overview engineers are ready to evaluate your needs and get your project started.

Compare

Why Automated Quality Control?

Flexibility

Case Studies

Calculator

learn about Node RED

Why Deep
Learning?

About Us

Culture and Life

Mission and Values

Team

join our
awesome
team

OV20i Use Case