Industrial camera in a precision visual inspection.
ARTIFICIAL VISION AND IMAGE PROCESSING

Machine Vision and Computer Graphics

Industrial platforms to capture, process and interpret images along with the process. We select compute, acceleration and interfaces based on the cameras, software and required response time.

Performance starts with the acquisition and processing architecture

Resolution, image cadence, number of cameras and model complexity determine the processing load. The choice must balance CPU, GPU, memory, storage and bandwidth without losing sight of the installation environment.

The architecture must also consider camera interfaces, synchronization, communication with PLCs and the platform life cycle. The goal is to process data locally and deliver useful results to the process with predictable latency.

Compute and platforms for industrial vision

Explore families for accelerated acquisition, development, and processing along the way.

Cinematic optical inspection station integrated into an industrial production line.

How we scale industrial vision systems

We start from the cameras, the software and the expected result to define processing, connectivity and operating conditions.

  • Quantify image flow Resolution, images per second, compression, and number of channels determine memory, storage, and bandwidth.
  • Select compute and acceleration CPU, GPU or accelerator are chosen depending on the algorithm, latency and thermal profile available.
  • Integrate with the process Camera interfaces, trigger signals, industrial Ethernet and communication with PLCs must be validated together.

Are you sizing a vision system?

Share cameras, software, response time and installation environment. The AGL team helps build a technical shortlist.

Talk to a technical specialist