Better Machine Vision in Low Light

Reliable identification of codes using the Ubicept solution

Modern perception systems can perform remarkable feats under ideal imaging conditions. However, when confronted with low light and fast motion, they often struggle to balance noise and blur.

To demonstrate this, we set up a shootout in our office. The test was simple: dim the lights, toss a box around, and see how an off-the-shelf QR code reader performs when provided with input from a low-light machine-vision camera (left) and the Ubicept solution (right):

In this still frame, you can see that both setups are able to identify the QR code, as indicated by the boxes drawn around them. Both setups perform well when the motion is slow, but the difference was clear when we began tossing the box back and forth:

Here, the machine vision camera failed to provide sufficient quality for identification while the Ubicept solution succeeded every single time! A closer look at the captured data reveals why:

The output from the machine vision camera was unusable because of noise and blur. We attempted to reduce the amount of blur by reducing the exposure time and increasing the gain, but this resulted in an even noisier image that was still unusable.

In contrast, the Ubicept solution provided incredibly clean data. You can even use your mobile phone to read the QR code from the above image!

To see more, check out our demos here. And if you’d like to learn more about how the Ubicept solution can help your machine vision system, please feel free to contact us any time!