The journey to becoming an enabling technology

There has been an enormous development in the past 25 years as vision technology evolved from using a disparate collection of components from other disciplines to the high performance dedicated systems available today. Today vision is very much considered to be an enabling technology. 

Vision integrated into a manufacturing process can have a significant effect in terms of reducing the amount of waste (defective product) produced, energy consumption required to produce the defective product and time wasted in producing defective product. End-of-line inspection, based on ‘pass/fail’ decisions can prevent defective product reaching the customer, but there could be a lot of waste. In principle, the earlier in the process that a vision system can be used, the earlier it would be possible to detect if the product is starting to go out of specification. This allows preventative action to be taken by adjusting the process with the objective of eliminating defective product altogether. 

Machine vision technology exists in a fast-moving environment, where developments in processing power, camera sensors (especially CMOS sensors), illumination, optics, software capabilities and image data handling standards constantly push back the boundaries of what is achievable. Cameras with higher resolution, faster operation and smaller physical size are linked with ever more powerful processing systems to enable more complex inspections to be carried out or give better performance for existing applications. 

Along with all of these developments has been the emergence of the ‘plug and play’ concept for vision which facilitates system design for vision systems integrators and opens up possibilities for non-specialist users to reap the benefits that vision inspection has to offer. This has opened up vision capabilities to machine builders who wish to integrate vision into their products and to end-users. While this has broadened the accessibility of vision, in terms of setting up a vision system, it should be remembered that a key requirement of getting good results is getting a good image to start with. Clearly, choosing a suitable camera is important, but getting the right illumination is critical, so some knowledge and understanding of vision is usually needed even for self-build systems. Vision systems integrators will be needed to supply complete turnkey solutions including product reconciliation, rejection and handling. 

There have been a number of key developments during the past 25 years that have contributed to plug and play capabilities. These include:

• Simple to use measurement tools

• Smart cameras

• Camera connectivity

• Embedded vision

Most of these developments are a consequence of the continuing evolution of semiconductor technology during this period which has allowed the development of smaller and faster processors, massive data storage and an extraordinary range of image sensors. 

An early stage in the development of vision technology was the use of PCs instead of minicomputers for image measurements. However, before the introduction of the PCI bus in 1993, no PC bus was fast enough to be used for image data transfer within a PC. The solution was typically to install processing power on the frame-grabber board or to use a secondary bus dedicated to image transfer. 

Although IBM-compatible PCs had begun to be used before 1995 as the basis for vision systems, the advent of Windows 95 made 'point and click' programming easier to implement. Improvements in PC programming power and memory paved the way for the development of PC-based image processing toolkits and libraries, which were used by vision experts to set up the inspection and measurement. From these came the self-contained machine vision applications with a simple interface framework that offer a plug and play approach for PC-based systems. These feature tools for a host of vision tasks such as detecting the presence/absence of features, part alignment, part measuring, surface inspection (looking for scratches, stains or irregular features), 1D and 2D code verification, character reading, label inspection etc. These then provide non-vision experts with easy-to-use tools to set up measurement routines.

The development of smart cameras represents a literal plug and play capability by moving all of the image processing capability into the camera itself. All of the measurements are made in the camera and a result output signal produced which can be sent directly to a PLC. In addition to built-in processors, the on-board software will feature the sort of measurement tools described above. Set-up can usually be made through a standard web browser. This can be reset for different product types and even moved to different locations to carry out different inspections. A different form of smart camera which has gained real traction in recent years is the 3D smart camera. These have truly been made possibility by the developments of processors capable of handling the 3D calculations but offer the possibility of individual or networked 3D imaging.

For PC-based vision systems, the images need to be transferred from the camera to the PC for analysis and measurement, so camera connectivity is very important. Smart cameras overcome this issue since the on-board measurement removes the need to transfer the actual images. However, image data transfer possibilities were transformed by the introduction of the GeniCam standard which is the umbrella standard required by all modern machine vision interface standards. It allows access to devices such as cameras or strobe controllers no matter which technology is used. The first dedicated data transmission standard for machine vision was CameraLink. Further standards, such as GigE Vision, CoaXPress, CameraLink HS and USB3 Vision have followed, offering different data transfer rates over a variety of distances to suit the particular application. These offer real plug-and-play capabilities since cameras from any manufacturer that meet a particular standard are interchangeable. 

Of these, GigE Vision has probably had the most impact in the factory environment since it allows data transfer over distances up to 100m using industry-standard Ethernet cables and switches. Consumer interfaces such as USB 2.0 and USB 3.0 and USB 3.1 also provide plug and play capabilities since cameras with these interfaces can be readily connected to a PC. However, these devices require the installation and use of the camera manufacturer’s own software rather than the generic software packages that are available for the machine vision standards.

The rapid evolution of computing power in embedded, single board computer systems is providing new, exciting possibilities for vision. Embedded vision systems based on platforms such as NVIDIA Jetson, Raspberry Pi, CompuLab and ODROID are finding increasing use in applications where space is constrained, cost is an issue and a self-contained vision solution is required. An even newer approach is the system on chip camera which is optimised for advanced digital imaging combined with a comprehensive image processing library (IPL). These systems, including smart cameras provide a scalable choice of embedded vision solutions.

Looking forward to the start of the next 25 years, machine vision is set to make a significant contribution to the smart factories of the future. Just as with current vision systems, connectivity between systems and the interchange of data will be crucial, especially as cameras generate large quantities of image data. To that end, in 2016, the German VDMA and the OPC Foundation set up the VDMA OPC Vision Initiative working group to develop an OPC UA companion specification for machine vision. This aims at straightforward integration of machine vision systems into production control and IT systems. At the lower level in smart factories, data will be produced by a host of smart sensors which will certainly include industrial cameras. Other applications could include bespoke manufacturing where vision systems are used to determine required dimensions which are immediately transferred into a manufacturing process. The move towards Industry 4.0 is already underway, with the emergence of products which can monitor an entire production process in real time, to identify which machines or processes have the greatest impact on overall productivity. Work is in progress involving some UKIVA members to integrate vision systems into this type of system. Code reading and verification, a well-established application for vision, will be a key factor in this process in order to eliminate operator data input error.