Defining the Merchant Machine Vision System Integrator Landscape

One of the more interesting aspects associated with the growth in the machine vision market has been the development of a system integrator infrastructure to support that growth. Back in the early 80’s, in the absence of that infrastructure the suppliers of the configurable machine vision products filled that need. However, although they might have conversant in machine vision technology, most had little or no application understanding. Since they were trying to establish themselves as viable players in the industry, many of these companies simply took on individual projects that had the promise to produce revenues.

Since these projects inevitably had significant engineering content and sometimes poorly defined needs with ensuing ‘‘creeping expectations,’‘ many of these suppliers of configurable machine vision products went out of business. Those companies that survived ultimately recognized that they had to become more product oriented and leave the more involved projects up to either merchant or captive (corporate in-house engineering groups) system integrators. The result was product companies that developed groups of offerings that were most often based on application-specific software, which eased the pain for the system integrator.

Back in 1984, there were fewer than 10 merchant system integrators and many of the Fortune 500 manufacturing companies had staff to perform integration. In many cases, internal staffing for such purposes was driven by proprietariness associated with production processes.

Over time, machine vision companies figured out what were industry-wide requirements and so the application software evolved. The underlying technology has evolved to include: far more computation power to handle algorithms that make applications more robust; price performance gained as machine vision industry piggy-backed on price-performance developments of camera and microprocessor technologies; and advances in lighting knowledge and lighting technology. The result is that the risk associated with undertaking a new machine vision project has decreased considerably over the years making it possible to succeed as a merchant system integrator.

Today, there are well over 600 merchant system integrators in North America alone. Many of these concentrate on specific niches associated with a family of complementary technologies: packaging lines, robot applications, distribution center applications, assembly line applications, etc. In other words, all merchant system integrators are not equal. Some rely on the suppliers of the configurable machine vision products they employ to provide the application software, which is applied to tried and true applications – OCR/OCV, 2D symbol reading, etc. In many of these cases, the application engineering is even ‘‘off-the-shelf.’‘ In other cases, integrators will use third-party software (or possibly even their own developed software) and engage in the application engineering to handle a unique application. For each application, it is important to find the right merchant system integrator.

These are good times for merchant system integrators as more companies outsource their requirements. A recent survey conducted by Control Engineering magazine, and reported on in their May 2007 issue, observes that there is increasing use of machine vision in manufacturing with the main frustrations related to setup and installation. As they point out, ‘‘not surprisingly, 30% of respondents use machine vision system integrators to help with installations, and 15% who do not already use system integrators plan to do so in the next year.’‘

To get a reading from the merchant system integrator perspective, we canvassed input for the following questions from many integrators. The following responded:

Perry Cornelius, Advanced System Consultant - ABCO Automation
David Dechow, President - Aptúra Machine Vision Systems
Brian Smithgall, President - Image Labs International
Louis DiCaire, Vice President - Orus Integration
Shelley Fellows, Vice President of Operations - Radix Controls

1. Can you briefly describe the services you provide as an independent merchant system integrator of machine vision systems? What other value-added services do you provide? What industries do you serve?

[Perry Cornelius – ABCO Automation] ABCO provides complete turnkey systems for machine vision, robotics and custom automation.  Our well-equipped machine vision lab allows for thorough and quick evaluations of customers’ samples.  ABCO’s internal engineering, project management, machining, fabrication, assembly and documentation departments allow for tight quality control and faster deliveries.  Some of the industries that ABCO serves are medical device, consumer goods, nuclear fuel, automotive, tobacco, beverage and appliance.

[David Dechow – Aptúra Machine Vision] We provide the following key services related exclusively to machine vision applications:

• Integrated Systems
• Design, implementation, installation and start-up of complete turnkey automation applications that involve machine vision technology. 

Machine vision solutions from Aptúra are fully guaranteed to perform to specifications.

• Engineering Services - Machine vision consulting, project management, programming, installation, and training – one-on-one or group classes – on- or off-site, specifically tailored to the customer’s equipment and needs.
• Technical Support - On-site service of existing machine vision automation equipment, regardless of the original manufacturer or integrator. Service may be contracted on a per-diem or fixed rate basis.  Long-term contracts and maintenance agreements are available.  Aptúra can provide emergency service, on-call 24 hours a day, through prior arrangement.

We serve a wide variety of industries including packaging, automotive, aerospace, food/beverage, pharmaceutical, electronics and medical device.

[Brian Smithgall – Image Labs International] We design and develop custom machine vision systems. We help people who are designing systems to identify the proper equipment for their projects. This equipment includes cameras, frame grabbers, lenses and software, for example. We serve all industries.

[Louis DiCaire -- Orus Integration] We are an integrator dedicated entirely to machine vision. Our services include complete turnkey systems to end-users, specific vision modules (hardware or software) to OEM partners and feasibility studies for projects where some elements need to be verified more deeply.

[Shelley Fellows - Radix Controls] We have provided customized vision solutions for robot guidance, traceability, quality and gauging since 1994.  We also provide automation design and build; controls design and programming; robot cell design and programming; custom software development design and programming; and system installation and start-up support.  We serve the automotive, food/beverage and pharmaceutical industries in North America.

2. What type of machine vision hardware do you normally integrate? Frame grabbers, smart cameras, embedded vision processors, etc.? And, why do you use that hardware?

[David]  We select hardware components based upon two principal criteria.  First, we must deliver to our customer a solution that works; and the hardware that we recommend must be able to perform the application.  Second, we respond to customer requests and specifications.  When possible (and usually it is possible), we will use the components specified by the customer or source of the customer opportunity.  Aptúra integrates smart cameras, open architecture cameras, and PC-based solutions with FireWire, GigE Vision™, Camera Link® or analog cameras.

[Brian] We often use Camera Link® cameras, frame grabbers and higher performance software. The jobs that can be done with smart cameras are better done with more local support. We occasionally use smart cameras as part of a larger system.

[Louis] Orus Integration is a one-stop shop in terms of an integrated vision system. It can go from a simple image acquisition system with little image processing capabilities, to a complete multi-camera, multi-server system for sorting items such as fruit, for example. Our niche is really in mid to high complexity projects that involve a lot of processing power, multi-cameras or motion devices like robotic or gantry systems. We've also developed a niche in fast laser 3D systems.

[Shelley] All of the above.  We recommend what we feel is the best hardware configuration for the task at hand, and have worked extensively with all of those types of vision platforms.  We have also developed our own real-time 3D vision tracking system in order to address an unfilled need in the market.

[Perry] If well-suited for an application, ABCO prefers smart cameras due to their ease-of-use, which shortens the development time for us and the learning curve for end-users.  For more challenging applications (e.g., very high speed, specific spectral range requirements, line scan, etc.), PC-based systems (i.e., frame grabbers, FireWire, USB, GigE Vision™, etc.) are used.

3. Has integrating machine vision become easier or harder over the years? Why? How has this impacted the price a customer pays?

[Brian] Prices have come down and performance has increased substantially.

[Louis] I think the hardware/software itself has become easier to use with more generic and more documented standards like GigE Vision™ and 1394 standards¬. For PC-based systems, more time was devoted to developing multi-standard platforms that can accept GigE Vision™, Camera Link®, LVDS, 1394, etc., in order to offer field-tested solutions that will stand the test of time. The price of complete integrated solutions has dropped for multi-camera systems or for really simple projects that smart cameras can handle. As a performance supplier (an integrator delivers a performance, not material), the long and complex part of any project is always to make sure the developed solution will suit the needs of the end-user -- not more, not less.

[Shelley] As customers become more familiar with the capabilities of machine vision, we are finding they are bringing increasingly more challenging vision requests to us.  Bin picking, high-speed acquisition and high-resolution surface scanning applications are all becoming more common.  As technology catches up with customer needs, the cost to install an application is dropping.  As an example, the cost of a vision-guided robotic application today is approximately one-third of what it was 10 years ago.

[Perry] It has become easier over the years.  One reason is a wider variety of off-the-shelf components (e.g., lighting, optics, smart cameras).   Due to lower hardware costs and simpler programming, the customer is paying less than in years past.

[David]  This is a more complex question than it might seem.  The difficulty of ‘‘integrating’‘ machine vision - the skill required to do any but the most trivial applications - has not changed much in 25+ years.  The specification of a correct and robust inspection and automation methodology has not become any easier.  One must have a certain depth of understanding to correctly apply lighting and optics.  One must understand the general inner workings of machine vision algorithms to know which will stand up to particular part variations and changes.  One must be able to predict the vagaries of part presentation and the effect that has on the imaging.  One must be well versed in the impact of pixel size and contrast for metrology and object recognition and differentiation.

That said, certain things have, of course, become easier.  There are more and more components specifically targeted to machine vision applications -- lighting, lenses, software, smart cameras.  It’s easier to find what one is looking for.  And thanks to more integrated development environments, for some applications, the development cycle has become shorter.  Menu-driven application configuration means that it takes a skilled integrator less time to make a smart camera do an inspection, so in some cases the integration is less expensive because it takes less time.  In other situations, however, the limitation of menu-based systems makes the job harder and causes more work and more expense as compared to a fully programmable environment.

4. Is the value added by merchant machine vision system integrators greater or less than it was five years ago? Why?

[Louis] Since smart cameras have boomed, two consequences appeared: 1- more end-users and non-vision specialists have access to this technology; and 2- machine vision systems have greater exposure in the industrial sector. For vision specialized integrators like Orus, the offer has to take into consideration the penetration of smart cameras in the market. For example, the value added by integrators is greater than ever since people know about the technology, have had small successes with small systems, and now see the potential for more complex vision system for process control, for example. That is where integrators like Orus have greatly added value by showing the full spectrum of solutions available to manufacturers (2D, 3D, X-Ray, line scan, area scan, thermography, etc.).

[Shelley] As smart cameras and their programming software become more user-friendly, engineers are finding they can install and program simple vision inspections without involving a third-party vision integrator.  However, for more challenging vision applications - multi camera, high speed, high resolution - or those in harsh or difficult environments, the machine vision system integrator is key to the success of the project.

[Perry] I think integrators are adding the same or more value today.   Most end-users that have little or no experience with machine vision welcome the expertise of an integrator.  As in years past, many end-users attempt to integrate their vision systems internally.  Some are successful, but integrators sometimes ‘‘finish’‘ many of them.

[David] The value is the same as it always has been.  The job of the machine vision integrator should be to mitigate risk in an inspection application.  Presumably, the expertise of the integrator means that the application will be successful.  Admittedly, the market price of this value has gone down just as hardware prices have gone down.  Further, the perceived value of integration services has been eroded because the barriers to entry into the business have been significantly reduced by customer perception.  It is easier now for one to say that they are a ‘‘machine vision integrator’‘ with a limited experience base than it ever was.

[Brian] The technology has evolved so end-users think they can do more of the jobs themselves, but they still need someone experienced with lighting and selecting the proper hardware.

5. What are the challenges in integrating machine vision these days?

[Shelley]  In a word - choice.  There are so many technologies that can be used to solve a particular vision problem that an end-user needs to rely on an expert in the field to assess their manufacturing challenge and recommend the optimal solution.  Also, as manufacturers maximize efficiency in their production operations, engineers and trades-people are faced with increasing demands on their time.  They need a system integrator who will invest the time to understand their operations and to design and build a vision solution that meets their needs.

[Perry] On the technical side, the core challenge of lighting is less (but still critical) than years ago due to a wider variety of off-the-shelf machine vision lighting.  However, software and communication issues still arise.  One example is a feature in the vision software that may not run properly because of some unknown service or program already running on the customer’s PC.  It can be time consuming resolving these types of issues.   That is, as the software gets user-friendlier, it also gets more complicated at the lower level.

[David] Beyond the obvious engineering challenges, one significant challenge is the prevailing customer perception that machine vision is a lot cheaper than it used be and that machine vision is so advanced that it can now do anything.  The former is only somewhat true; the latter is clearly untrue.  Machine vision is, to an extent, easier to use, hardware is faster than ever, resolutions are higher and algorithms in some cases are more complex.  But for the most part, what could not be done 10 years ago because of imaging, lighting, or automation constraints can not be successfully done today.

[Louis] We think the challenge has not really changed and will stay the same for the coming years: make sure the system will answer the needs of the end-user in terms of performance, reliability and versatility. Each technical solution has pros and cons, so the role of the integrator is to expose these pros and cons to the end user to make sure the final decision was made in a perceptive way. The end user must understand that the integrator is not a distributor, so there is no advantage to pushing a product or a technology if it's not suited to that application’s needs.

6. What is the role of the customer in a machine vision integration project?

[Perry] The role of the customer is to initially provide a detailed specification of their expectations along with a wide variety of good, bad and borderline samples.  It is also their responsibility to take ownership of the project, which requires that maintenance personnel and operators become involved in the early stages of the project and become properly trained.

[David] The customer must provide accurate specifications about their parts, required inspections and automation processes.  They must have a ‘‘champion’‘ who will understand the machine vision application and ultimately be able to support it.

[Brian] They need to find integration help that has the right experience for the project. They should define clear objectives and monitor progress.

[Louis] The role of the customer is crucial. The customer is the premier information provider for a project, especially in the design phase where the whole system architecture takes place. The more accurate the information is, the more suited the system will be.

[Shelley]  In the initial stage of designing and estimating a vision solution, the customer needs to be the key communicator of project requirements (both current and future).  Without full disclosure of operating conditions, manufacturing processes and vision tolerances, the success of a vision solution can be compromised.

Once a project is underway, the customer also needs to work closely with the vision system integrator to review designs, to complete any internal changes to processes or equipment and to educate their project team as to what will be necessary for a successful project.  The process is a collaborative project from start to finish.

7. What is the role of the vision hardware supplier in a machine vision integration project?

[David] The hardware supplier must deliver a competent, operationally sound product, generally free of bugs, and must be able to and willing to support the product from a hardware/software problem perspective.

[Brian] Provide reliable solutions that work as described.

[Louis] With many vision systems, the performance is specified but not the material required to achieve that level of performance. We see the vision hardware suppliers as partners in order to provide support and especially demonstration units so we can try the selected hardware and see how they work together to make a complete system coherent. If a light does not have enough intensity, we typically want to know as soon as possible and be presented with alternatives.

[Shelley] Suppliers of cameras, frame grabbers, lighting and optics alike need to work closely with the system integrator to understand customer requirements and ensure that the equipment they are supplying is the right choice for the job.  Of course, we rely on them for high quality, well- engineered products that will survive in challenging manufacturing environments.

[Perry] The role of the hardware supplier is to provide support for their products when needed (e.g., provide updates on their latest features, resolve bug issues, provide evaluation units and/or sales support, etc.).

8. What is the role of the system integrator?

[Brian] Help the customer achieve their goals.

[Louis] The role of a system integrator is to be knowledgeable about different vision system architectures, expose the different scenarios to the client, highlight the performances of each proposed system, highlight the limitations of each system, design, build, test, install and optimize the system.

[Shelley] Three E's:

• Education - during the proposal design phase, during project execution and following installation, the system integrator needs to educate the customer 
• Expertise - system integrators must be very knowledgeable about the products they support, investing in ongoing training and education of their technical/sales staff
• Excellence - the system integrator must provide a high quality solution to meet the customer's needs (not just features, but also budget and timeline), which necessitates an attitude of continuous improvement with respect to the system integrator's operations and a commitment to a well engineered and designed solution for all their customers.

[Perry] The integrator is responsible for the success of the project, which means providing the customer with what was promised (e.g., design, installation, programming, training, documentation, validation, etc.).   Effective communication plays a big role.  It is the integrator’s responsibility to request the appropriate information from the customer and to notify the customer immediately if issues arise.

[David] The integrator must make the system work, period.  The integrator must take sole responsibility for the system. The integrator must work with the hardware supplier and not whine about hardware/software bugs.

9. What are common mistakes made by buyers of machine vision systems?

[Louis] Buying brand over technology. There are fundamental differences between types of vision systems that can turn a potentially successful project into a nightmare. Of course, if there are only subtle variations between two solutions, only in these situations one could favor a specific brand if he/she is used to the software tools. For all other situations, if a line scan is required by the project and the brand a customer likes is not offering a line scan, it would be a mistake to try to arrange something with area scan. Same thing is applied to 3D versus 2D. Let the technology be the first filter.

[Shelley] The single most common mistake is to omit funding for training.  In particular, a plant that is new to vision must prepare to invest in the training of operators, trades and engineers.  Without the understanding of how to interact with and maintain the equipment, the probability of failure for the vision station following completion of installation is much higher than if training is included in the project.

The second most common mistake is one that the machine vision industry can help with - lack of understanding of what is required for a successful vision project.  Budget, timelines, personnel requirements - customers need to be educated about realistic requirements for a successful project.

[Perry] One mistake is not having the appropriate personnel trained on the system.  If the day-to-day operators of the process are comfortable and familiar with the system, then potential issues can be avoided.

[David] Having unrealistic expectations about performance, capability, and pricing.  Specifying components and not solutions.  Failing to adequately communicate requirements, specifications and needs.

[Brian] Not defining clear goals, not monitoring progress, not understanding limits of the technology and changing their minds about goals.

10. What advice would you give to a prospective customer for a machine vision system integration project?

[Shelley] Choose a system integrator that you can have confidence in, who has demonstrated expertise in your type of vision project and industry - ask for references and examples of similar projects.

[Perry] They should carefully develop a detailed specification on what is actually needed.  Do not require the system to find every defect possible; try to stick to the inspections that will give the greatest return on investment.  It they cannot or do not want to integrate the project internally, then they should meet with and evaluate several different integrators before making a decision.  Some integrators may have more experience with their product and processes than others.

[David] Specify, specify, specify your application.  If necessary, work with an integrator or consultant to prepare a clear and detailed specification of the application.

[Louis] Make sure you have a system champion in your team. Make sure the whole hierarchy is involved, from the plant manager to the operator on the factory-floor and that you get input from all levels. Sometimes, small details can make the difference between success and a failure -employee acceptance of the new system is one of them.

11. What are the overall challenges a machine vision integrator faces today to make a successful business out of integration?

[Perry] Controlling ‘‘scope creep’‘.  That is, try to deliver only what was agreed upon.  Trying to accommodate all the ‘‘by the way’‘ requests from a customer can quickly drain the profits.  This is good for the customer in the long term, also.  It is much easier for a profitable integrator to provide long-term support to a customer than one that is struggling financially.

[David]

• Unrealistic expectations by the customer regarding the scope of work required to implement machine vision technology.
• The pervasive ‘‘do it yourself’‘ mind-set promoted by hardware suppliers; even to the detriment of a customer’s project.
• So-called ‘‘free’‘ integration and application programming by hardware suppliers and distributors – usually needing to be corrected later by a qualified integrator.
• High costs of staffing, travel, etc.

[Brian] Getting enough experience.

[Shelley] Our biggest challenge is to keep up with trends in emerging vision technologies.

12. What changes in the underlying machine vision technology do you anticipate in the next 2 – 3 years? Will these make things easier for a machine vision system integrator? How will this affect the ultimate price for a customer?

[David] My very radical prediction is that general-purpose machine vision will slowly and slightly begin to migrate back to PC-based systems.  This architecture has great value and price/performance ratio.  End-users may begin to embrace this form of machine vision in lieu of smart cameras.  I’m not predicting the demise of the smart camera, just perhaps a slight shift.

[Brian] Less frame grabbers, performance improvements, ability to do tougher jobs more reliably. Prices will be lower for comparable jobs.

[Louis] GigE Vision™ is surely a standard that will simplify system architecture, simplify the electrical deployment of the solution (less bulky cables) and make the vision system more web-enabled (control the system with your web browser). As for the price, in the mid to high complexity system, the engineering part of the complete solution (design, test, optimization) will not vary a lot even though the components will drop in price. Multi-core processing systems will also simplify the deployment of PC-based systems.     

[Shelley]  Vision cameras will become more powerful, smaller and more cost effective.

[Perry] Continued increase in performance with price reductions and more features will be added to the software.  In turn, applications not previously considered may become feasible.  The ‘‘low’‘ risk vision applications will become less expensive, but the new applications that require state-of-the-art hardware and software will be more expensive than typical applications (at least for a while).