Vision systems increase performance, lower cost, get easier

Machine vision systems relying on more open architectures are easier to incorporate into more applications, offer higher levels of performance, and are more economical to use, according to recent Control Engineering research, backed up with comments from those working with and developing the technologies.



Machine vision systems relying on more open architectures are easier to incorporate into more applications, offer higher levels of performance, and are more economical to use, according to recent Control Engineering research, backed up with comments from those working with and developing the technologies. (See also “ Product Research: Machine Vision Looks Well Beyond Inspection .”)


Fernando Ventresca, automation manager, Shelley Industrial Automation, DVT Automation Solution Provider, said, “A pressing quality control issue surrounding vision today is on error proofing ability at the “source” rather than error-proofing a final product. It does not make fiscal sense to value-add into a product that inevitably is defective. Machine vision is being applied during the process… at the source rather than at the end of the production process.” (Cognex recently acquired DVT, noting the value of its technologies and distributors. For more, see “ Cognex buys DVT for $115 million .”)


Machine vision has found growth and success in areas where the applications are simplistic, such as with sensor replacement, Ventresca said. “More people are finding a need for machine vision because of the cost vs. performance. Years ago, the price of machine vision far outweighed its capability. Today that trend has reversed. Therefore, the applications that best suit machine vision today are fairly simplistic ones, presence/absence, product verification, correct assembly, and orientation cover a far greater spectrum. In regards to the future: More success needs an approval from robotic hand-shaking, coordinate transformation, and 2D and 3D extrapolating. Many inroads have been made but more energy and effort are required. Experience, education/training, and partners willing to go that extra mile have been our key to success thus far. Many companies have built an empire on machine vision. A vision system can be the cornerstone to a complete turnkey pick-and-place, orient/measure, pass-fail piece of equipment that can make or break a contract for an automotive parts supplier or automaker. Partnering with good integrators who understand the automotive mentality and are willing to educate and train themselves on your product is a big factor.”


William J. Amato, president, Phoenix Automation Group Inc., added, “Areas of development still lie in the bar code reading algorithms; however, DVT has made incredible strides with its new Intellution Set-up Software. Having an open architecture for implementation into most plant-floor control systems, DVT is positioned to be more than just an “island on the plant-floor." This allows the DVT system to seamlessly connect to the many controls and mediums that the automo-tive world demands, such as operator interface, SPC software, as well as robotic systems. Phoenix Automation Group and its customers have experienced great successes with these value-bundled solutions.”


Mark Sippel, In-Sight Principal Product Marketing Manager for In-Sight vision sensors at Cognex, observed, “As the costs associated with deploying vision come down, and as vision gets easier to use, more opportunities open up for it. As a result, use of vision has increased for in-plant applications directly due to these factors, along with greater exposure and understanding of the benefits of vision for improving quality and the manufacturing processes. With the need to gather and share data from devices in plants increasing, Ethernet is quickly becoming the communication method of choice. The capability of supporting easy access and use of several Ethernet protocols is necessary to maintaining user flexibility in new applications.”


Kyle Voosen, machine vision product manager, National Instruments, had some views about networking for machine vision. “If you look to the dominant industrial buses, Ethernet is certainly taking over as a hardware medium: EtherNet/IP, Modbus/TCP, ProfiNet. I would argue that just because a vision sensor has an Ethernet port on it doesn't mean it can speak to an AB PLC. The software must speak the correct protocol. Most smart camera makers still require an external gateway to convert data to other communication protocols.”


Commenting on product selection criteria favored by Control Engineering subscribers responding to a recent machine vision survey, Voosen said, “that accuracy, not speed, is the biggest issue concerning machine vision users. This means that it is not the hardware speed that is holding ma-chine vision products back, but the software accuracy and repeatability.” National Instruments provides several avenues to expand our configuration software, Vision Builder AI, Voosen said. The software can call custom steps or code written in a programming language, so if there’s a need to insert an algorithm not included in the software, you can do it without having to port the entire project over to a programming language. Another option is to generate LabView code from Vision Builder AI, he said. Based on survey results, “It would seem that machine vision hardware has achieved the speeds and form factors necessary for most applications, but that challenges remain when it comes to software accuracy and ease-of-use,” he said.


Joshua Jelonek, machine vision application engineer, Keyence Corp. of America, said “Every passing year not only brings reductions in the costs associated with installing a machine vision system, but also offers systems with intuitive user interface's that don't sacrifice power and flexibility. Performance should always be number one; if the system can't perform the desired task then there is no point in purchasing it. However, technical support is also a very important component to successfully implementing a machine vision system, and it's not surprising that ease of use is rated next. The two go hand and hand. If the system is simple to set up and program, inte-gration time will be reduced and the need for technical support is not great.”


Robert Lee, vision product marketing manager, Omron Electronics LLC, said, “Over the last two years, there has been a notable market trend in the use and proliferation of vision sensors. In comparison to 2004 statistics, while there has been a decrease in demand in the OEM resale sector, demand in vision sensors has been most notable by users for both the OEM (resale) sector and for inplant requirements. This could also imply that the thirst for vision sensors has been par-ticularly exacted by integrators to satisfy the glut for these products at the expense of direct distribution through OEMs and in-plant use. Several attributes can be derived from this, the most significant being that emphasis for the commissioning of these devices is increasingly and more importantly being supported through integration, with less focus placed on the end user and OEM distribution. This development is in many respects fueled by increasing demand without objec-tions from capital budgetary restraints. However, market acceptance of these products is still reduced because of priority relative to other automation products, understanding of the underlying technology, and industry acceptance. In comparison, difficulties of use and allocation of engineering resources did not serve to impose any impediments on these products.


“Communication features also represent a significance in this scenario,” Lee continued, “with an escalating propagation in Ethernet technology at the expense of RS-232 technology. These identi-cal characteristics epitomize customers’ criteria for product selection for this equipment: simplic-ity, technology, product features, etc. In a most recent survey which concords with market developments in vision sensor technology, it can be observed that the primary applications for these products are inspection, bar code reading, motion control, gauging and for disposition in robotic handling. Prominence of this technology in packaging machines, product manufacturing, material handling, printing and character recognition is less pronounced, and may betray areas that are of infantile growth since motion control, continuous processing, verification and diagnostics still represent areas of the largest growth. In summation, performance, technical support, and com-plete automation solutions encompassing machine vision technology will serve to differentiate the most significant participants in this arena.”


Ben Dawson, director of stragetic development, “ipd,” a division of Coreco Imaging Inc. noted that performance is the top criteria for a machine vision product. (Coreco was also recently ac-quired; see “ Dalsa purchasing Coreco for about $72 million ” for more.) “This includes speed, accuracy and preci-sion, and the ability to do the vision task in less than ideal environments and with less than per-fect parts; 56% of respondents also listed‘Ruggedness’ as a choice factor, which also suggests that systems are being deployed in more demanding environments. Speed, however, was a factor for only 38% of respondents. My interpretation is that vision companies are successful in moving machine vision to the factory floor and are meeting the need for speed. This is supported by having only 10% of respondents say ‘Acceptance by factory personnel’ is an issue.


“However,” Dawson added, “this also suggests that robustness, ease of use (67%), and ease of setup (63%) need work; ipd provides very easy-to-use machine vision. For example, our products use terms that are familiar to users—measurement, tolerance, etc.—and hide nasty details. Looking forward, performance and ease-of-use will increase and this will enlarge the range of applica-tions for machine vision. Increasing the range of parts that can be inspected is a major challenge. Most machine vision systems are currently limited to well-defined parts that are carefully illuminated and presented to the vision system. However, many products are either presented in poorly controlled environments or are have high variability, such as with food products. I think that developing robust algorithms that can deal with more variability is the major challenge—and opportunity—in machine vision.


—Mark T. Hoske, editor-in-chief, Control Engineering,


click me