Simply put, machine vision is the ability of an electronic system to see. Advanced processing systems analyze images to make a decision or classification. Automated product inspection, face recognition, and collision avoidance technology are possible thanks to machine vision technology. Machine vision cameras must operate with total reliability all day, every day, for many years, even in tough environments where there might be constant vibrations.

Machine vision cameras come with many features to manipulate and create a specific type of image. Basic and advanced features in machine vision include:

Ability to define a narrow field of view and ignore the rest of the scene
Automatic color correction
Timestamps to synchronize images to readings from other sensors
Synchronizing image capture with signals from other hardware devices
Synchronize image capture for multiple cameras
Precise control of camera configuration with in-house software development kit
Easy integration with third party machine vision software

Machine vision enables electronic and mechanical systems to turn visual information into decisions. Developers can take this a step further and use deep learning to quickly automate complex and subjective decision making. Deep learning mimics how the human brain processes data, using neural networks to distinguish between noteworthy anomalies vs. natural variations. The result is the ability to analyze more complex patterns, develop systems faster, deliver higher quality products, and enhance productivity.

Normally, deep learning systems require separate cameras and computer systems. However, advanced software systems allow integrators to deploy trained neural networks directly to the camera. This reduces system cost and complexity by enabling decisions to be made directly on-camera, in many cases without a host PC.

Application Example

Automation and machine vision technology that rely on infrared (IR) sensors are critical to a wide range of industries—from aerospace and automotive to cell phone production. Incorporating IR sensors usually involves rigorous testing to make sure the data that’s being collected is processed correctly. While it can be relatively easy to run multiple trials on cars and phones, a guided missile in flight 5000 miles away doesn’t allow for repeated field testing, so errors that aren’t caught early can lead to monumental costs later. Scenarios like these are where Chip Design Systems (CDS) come into play.

CDS creates IR projectors that produce scenes made of IR light to run tests and simulations for IR sensors—sort of like a VR headset for machines. CDS’s primary clients are government agencies and the projectors they create for these contracts must reach an astounding level of accuracy to meet their customers’ needs: often having to produce simulations for targets moving at supersonic speeds. To give a comparison, the average consumer monitor projects an image around 60 to 120 hertz; CDS’s projectors can display scenes at a rate of 50,000 hertz. On top of the high framerate, CDS can simulate temperatures exceeding 1000 Kelvin and at a resolution up to 2000 × 2000.

*Actual projected IR imagery of a parrot from a CDS IR Scene Projector*

The capabilities of CDS’s IR projection are clear , but testing them poses a challenge, as neither the human eye nor non-military grade cameras can capture all the data the projectors emit. To make sure the projectors work properly, CDS needed something capable of capturing as much IR information as their projectors can emit. The solution is high speed, high-definition infrared science cameras from Teledyne FLIR (i.e. X-Series models).

CDS frequently runs tests inside their lab to make sure the emitters are working correctly and push performance further. They perform tests by carefully aligning the FLIR camera with the emitter to capture the light and then, through the use of FLIR research application software (ex. Research Studio), control the camera settings to optimize the capture and display the resulting imagery. To guarantee repeatability, CDS has also developed its own code by leveraging the FLIR Science Camera SDK to automate the testing process and avoid human error from camera operation.

CDS testing one of their projector systems with a FLIR SC8200. This specific projector system uses liquid nitrogen to perform at optimal levels. An equivalent FLIR model would be the FLIR X8580, a cooled midwave HD camera that can reach frame rates of 181 Hz.

“With the help of FLIR products, we are able to demonstrate HD resolutions, hypersonic frames rates, and hot apparent temperature of our IR projector system to our end users and customers,” says Fouad Kiamilev, CTO of Chip Design Systems.

One feature that particularly stood out to CDS was the camera’s ability to perform non-uniformity correction, NUC for short, on the images from their projectors. Performing NUC in the FLIR camera helps CDS catch the occasional blemish in a scene or a faulty emitter. With the combination of the camera’s high resolution and non-uniformity correction, CDS can check every pixel in a scene’s image to verify their projector is emitting accurately.

A Teledyne FLIR customer since 2010, Chip Design Systems currently owns four IR cameras in their lab with plans to implement more in the future. Kiamilev says that CDS is aiming to drive projector performance even further in the future with higher resolutions, faster frame rates, and lower costs.