Machine vision systems - a view from the top

Visual inspection is an increasingly important aspect of modern industrial processes. Whether it is used for assisting in sorting bulk materials in pharmaceutical processing, inspecting the quality of produce in food manufacturing or informing the positioning of a delta robot in packaging, machine vision (MV) significantly increases the efficiency and productivity of various industrial processes. It's for this reason that the latest figures from MarketsandMarkets estimate the global MV market to be valued at $9.6 billion in 2020, with it expected to grow to $13bn by 2025. Even this figure is likely overshadowed by the value it adds to industrial businesses annually, by virtue of increasing throughput, minimising faulty products and reducing labour time for inspection tasks.

However, there remains one consistent challenge with MV systems, which is the ease of configuration and processing. This, in turn, affects the scalability. Typically, MV systems are connected to a specialised, high-performance industrial computer that is separate to a plant's other automation systems. A setup like this means that specialist knowledge is needed to adjust system parameters and configure cameras, in an environment that is often unfamiliar to many automation engineers. The specialism required presents not only a limitation to the scalability of MV systems in terms of the availability of skills to programme and manage the processing system itself, but also a barrier to true efficiency.

For example, imagine a food processing plant where one vision inspection system is required for raw material sorting, another for raw material defect identification, a third MV system for quality assurance post-processing and finally an MV-equipped robotic system for packaging. In this scenario, a change to the production process would require quick adjustment of multiple systems, and any issues would need to be addressed by a small number of skilled technical staff. For large operations, it's a high cost option that simply isn't scalable.

There is also the matter of latency. If the image processing is completed on a separate system to the motion control and automation, then that data needs to be sent to the relevant systems, where it is actioned accordingly. The logical solution is to integrate MV processing and management into the same system as the motion control and automation, to increase responsiveness and to support more engineers to adjust the vision processing software. This is what Beckhoff has recently done with the launch of TwinCAT Vision, an extension to our TwinCAT 3 PC-based control software.

It brings image processing into a single platform alongside programmable logic controllers (PLCs), human-machine interface (HMI), motion control and high-end measurement technology. With everything all in one platform, machines can respond to vision input data in real time, eliminating delays in motion systems. The PLC environment also means that the system uses PLC programming languages and the same configuration tools as used for fieldbuses. This means that adjusting MV systems is an easier task for automation engineers, allowing easy configuration and calibration of cameras, as well as supporting an instant review of any changes made.

In addition, TwinCAT Vision builds on the Beckhoff open control technology philosophy. The system is hardware-neutral, meaning it works with both line-scan and area-scan cameras via a GigE Vision interface. This makes it applicable to all manner of MV applications. MV systems have come a long way from the optical sorting systems of the 1930s, with advanced functionality and increased importance. Just as the physical systems continue to develop, it's important we also look at new ways of advancing processing software to ensure these systems continue to provide efficiency for years to come.

Learn more at www.beckhoff.co.uk.