2015 Will Be a Watershed Year

April 2015

Sony Stopping Production of CCD Sensors in 2017

The company that pioneered and greatly improved CCD technology is transitioning to CMOS. For those stalwarts still specifying CCD sensors, this is a not-so-subtle hint. It’s time to move on.

Lots of New CMOS Image Sensors

Sony CMOS Image Sensors

The Sony Exmore sensors have truly impressive specifications. Amazing quantum efficiency, low temporal dark noise, and high saturation capacity make for an excellent dynamic range. By all these benchmarks, Exmore sensors compare favorably against most CCD sensors.

There are also four new sensors coming from OnSemi. These sensors, when combined with a USB3 Vision interface, will challenge high-cost cameras built for moderately high-speed imaging. Inexpensive cameras will soon be delivering up to 850 images per second.

Our array of vision solutions will include both Sony Exmore and OnSemi Python sensors by end of the third quarter.

New Intel Microarchitecture

Intel Skylake

Code named "Skylake," Intel's new microarchitecture is the next advancement in mainstream computing. Media reported earlier this year that Intel considers Skylake its "most significant processor" for a decade, due to its enhanced power efficiency and wireless capabilities. And its processing power may come in handy if you’re contemplating analysis of 850 images a second from multiple cameras. Our systems will be available with Skylake processors by end of year.

Big Data Comes to Manufacturing

It might have been predicted. The biggest thing to happen in machine vision in a very long time came from outside the industrial sector. More specifically, it came from the University of Toronto in the form of their SuperVision deep convolutional neural network algorithm.

Since the SuperVision algorithm was developed in 2012, a team of Google engineers has improved it, making it nearly as accurate as humans. Imagine the value of automatically identifying the content of images on Facebook, Instagram and Pinterest. And now, in 2015, we’re starting to see these deep learning techniques applied to industrial manufacturing. This is bleeding edge stuff not for the faint of heart. But in time, we’ll look back and see this was the year it began.

But What About the Installed Base?

Sensors becoming obsolete forces cameras to become obsolete. CPUs becoming obsolete forces computers to become obsolete. We can’t freeze technology to protect your previous investments, but we can manage it.

We design vision systems using a modular approach. Each major component uses a stable, standard interface to communicate with other components. For example, our fourth generation hardware uses GigE Vision (Ethernet) to communicate with cameras and USB to communicate with discrete I/O. The idea is that when a component fails and is no longer available, it can be replaced with a similar component having the same interface. There is no need to discard the entire vision system. This approach protects your technology investment, while also enabling economical system upgrades.

We’re only four months into 2015. Imagine what might happen before year end. Someday we’ll look back on 2015 as a watershed year for machine vision.