Predicting the future, this isn’t. Let’s have a look at the state of the technology and extrapolate some trends.

Fewer shades of grey

At the moment we see monochrome (grey-scale) machine vision cameras in use across the industry for their current benefits in both resolution and price.

However, with the increasing modularisation and commoditisation of image processing, optics and lighting, colour is becoming a viable option in even the most demanding of machine vision applications.

Being able to construct modular LED light systems to complement astute choices of filters, lenses and processing equipment looks likely to lead to a competitive advantage, particularly in detection and recognition applications.

Working smarter 

Smart cameras currently account for a large proportion of the machine vision system growth rate.

This is largely driven by the continuing decrease in size, and increase in processing power, of the electronics involved.

With the standard footprint of camera units relatively fixed, the trend is likely to continue towards smarter, rather than smaller, units. For all the reasons we outlined in our previous article, it is probable that smart cameras will continue their penetration into application which traditionally have required the massive processing power of dedicated servers.

Adaptation to this processing ability may well take the form of pushing automated decision making into the fully automatic arena; more on that, later.

Towards greater utility 

Given the trend towards cheaper, more versatile machine vision components (smart cameras, modular LED lighting, et cetera) reducing the redundancy inherent in dedicated, specialised systems seems almost inevitable.

If relatively simple changes in lighting (turn off the blue LED ring light, turn on the red LED backlight) and launching a new image processing program on the smart cameras of a particular production line, from a single desktop PC, makes that line available for a different product, then there is considerable scope for expanding multi-use lines. That reduces the need for having lines quiescent while the plant changes production focus.

Looking deeper

Another technology on the rise is 3D imaging.

A competent 3D machine vision system, whether using stereo vision, point cloud techniques, 3D triangulation, or some other method can deal with variations in roll, pitch and yaw angle of the pixels making up the image of a product on a conveyor belt, given the expected dimensions.

The processing power available with modern computing systems allows for real-time image processing which may make it possible to improve current systems with 3D machine vision.

Computing within

The most common operating system, currently, for image processing systems and smart cameras alike is Windows. However, with Linux (or system variants like Android) continually growing in both industrial and embedded computing platforms (like mobile phones and DVR units), the likelihood is that Linux-based machine vision devices and applications will become more pervasive.

Another subject of growth, within the area of computing, is the user interface. There is rapid development in terms of both coding and UX design. “Clean room” and “wash-down” environments may drive the desire for, and development of, “hands-off” interfaces, where visually-aided programming from a standard subset of plugins could allow for machine vision programming via a machine vision system…

Feedback loops

Machine vision systems have migrated between industrial uses and other arenas.

Sometimes the uses to which machine vision have been surprising, but, once done, obviously inevitable. It sometimes also seems that this is a one-way street, where industrial innovation with machine vision benefits the wider world. As this use has matured, though, feedback loops are bringing outside innovation back into industrial machine vision.

• From ‘uncontrolled’ environments into industry: this is potentially the most exciting of the innovations making their way back into the industrial milieu.
• Interactive gaming demands that commodity machine vision is able to deal with random input. Although it is not critical (except to the player), it is necessary that it is near-real-time in reaction.
• Security and traffic control, in particular, is another driver for high resolution imaging. The proliferation of CCTV and other security cameras is pushing the need for low-cost, high resolution devices.
• Self-drive cars are also driving (excuse the pun) machine vision innovation, especially since Google entered the arena with ambitions for driverless vehicles operating in the public space. If ever there was a need for real-time, 3D machine vision, it is to stop a driverless car hitting a jay-walking child. Such systems must also be able to deal with ambient, uncontrolled lighting. In the shorter term, certain high-end cars include colour machine vision cameras in their safety repertoire.
• Other areas such as robot warehouse management and agricultural pickers, graders and sorters are re-using industrial machine vision systems, but expanding the remit. Those new requirements for colour and irregular shape matching are feeding back to industry.
• Universities on learning algorithms and interfaces: machine learning and robotics programmes within higher acadaemic centres have been fed by the advances in machine vision. They have, in turn, produced more intricate and efficient algorithms which make their way back to practical uses. More intuitive interfaces, should they arise, are likely to come from acadaemia.
• IT industry & commodity components: the machine vision systems used in industry have made demands of the IT industry in terms of hardware and software. The IT industry has responded by making available faster, cheaper and more accessible components, such as faster, multi-core CPUs, parallelisation across GPUs and more intuitive control programs, embedded Linux. Mobile phones, especially smart-phones, have played their part, too: android, low-power mobile processors, and commoditised digital camera parts.

Last words 

“All of life can be broken down into moments of transition or moments of revelation. This had the feeling of both.” – G’kar (Babylon 5, “Z’ha’dum”)

At present, we are looking at a moment of transition in the industry as the evolution of the technologies, already in play, continues. If we see the emergence of a disruptive technology able to greatly facilitate or obviate any of the areas mentioned above, we could find ourselves in a moment of revelation.
It certainly does have the feeling of both, though.