• Font Size:
  • A
  • A
  • A

Feature Articles

Reducing Risks in Deploying a Machine Vision System. . . . by Nello Zuech, President Vision Systems International, Consultancy

by Nello Zuech, Contributing Editor - AIA

The good news when it comes to applying machine vision today is that many applications have already been addressed. In some cases the application has resulted in a ‘‘canned’‘ solution including software, man-machine interface, ‘‘staging’‘ (camera, lighting, optics arrangement) and even line integration tools. Some of these ‘‘canned’‘ solutions include optical character recognition/verification, 1D and 2D symbol reading, filled bottle/label inspection, alignment/fiducial finding, semiconductor package inspection, LCD/LED inspection, fiber optic inspection, closure inspection, etc. Given these are ‘‘canned’‘ the risk to deployment is low, the non-recurring engineering nominal and the overall cost often very reasonable. So reasonable that it would be unreasonable not to deploy the machine vision system given the quality improvement that will ensue.

The purchase of such systems still requires developing some familiarity with machine vision technology as well as the vendors. Vendors offering these solutions generally offer various levels of training on the subject of machine vision. In addition one can find tutorials at various conferences including those sponsored by the Automated Imaging Association. In addition to developing some level of understanding of the fundamentals of machine vision, even though it looks like a ‘‘canned’‘ system will satisfy the requirements, it is still important to prepare a document defining the requirements.

Since all ‘‘canned’‘ systems ostensibly addressing the same application do not address the application the same way, their performance will vary somewhat. By defining your requirements it will be possible to measure the performance of each of the products from each of the vendors being considered against specific criteria/parameters. It is even possible that none of the systems addressing your ‘‘canned’‘ application will fully satisfy your requirements as defined because of the variables associated with your specific manufacturing process. In this case it would be important to identify the product with the most flexibility that, when fine-tuned, will yield performance consistent with your requirements.

Today it is also true that the underlying technologies for machine vision, such as cameras, lighting and optics, have been ‘‘productized.’‘ What used to be a special design can often be found in a catalog. The net result is that non-recurring engineering is much less than it might have been several years ago even when it comes to addressing applications that have a unique character to them or that require customization.

Nevertheless, when pursuing applications that require engineering it is wise to do so using a systematic process to assure that you will be satisfied with the system that is ultimately delivered. The first step is to assess the machine vision vendor community to determine the subset of vendors to consider for your project. There are different kinds of machine vision vendors and depending on the application, or the extent of the engineering that one is prepared to undertake, one or another type may be more appropriate to be on a bidder’s list.

For example, virtually every industry has ‘‘generic’‘ applications – applications that one will find in virtually all companies in that industry. These include pattern and unpatterned wafer inspection and 3D co-planarity interconnect measuring and package inspection systems in the semiconductor industry; bareboard, solder paste and assemble board inspection systems for the electronic industry; plastic, metal and glass container inspection systems for both the container manufacturer and filler, solid dosage and vial/ampoule inspection systems for the pharmaceutical industry; 3D-based volume measuring systems, mark sensing and grading systems for the forest products industry; systems for both fresh pack and food processors; systems for registration control and inspection in the print industry; etc.

There are also application specific machine vision systems, such as web scanners that address applications where products are produced in continuous sheets, like primary metals, plastic, textiles, nonwovens, film, foils, etc. There are also X-ray-based application-specific machine vision systems for the tire and food industry as well as for the solder joint application in the electronic industry and package inspection systems for the semiconductor industry.

In any event when you have one of these applications it is best to identify those companies that have been offering the specific system for some time. Inevitably the performance of those systems will reflect an understanding of the application variables. Be aware the performance of all the systems addressing a given application will not be identical because of their respective designs, algorithms and interpretations of the application. Consequently, although one is purchasing what might appear to be a catalog item, it is still important to have first developed a functional specification outlining the performance expected in the system you purchase. This document should reflect your interpretation of the requirements based on your own experiences within your operations.

Having so defined the requirements makes it easier to compare the performance of the proposed systems in terms of how closely each of the systems satisfies your own perception of the requirements. It is always best to seek out these suppliers when your application fits into one the application-specific machine vision classes because ultimately the risk of failure will be lower. These companies will have already been around the horn, so to speak. Contracting with a company that proposes to deliver one of these application-specific machine vision systems for the first time will increase the risk of failure. Inevitably, the company will not be familiar with all the variables associated with the application leading to delivery of a system that may perform marginally.

More often than not suppliers of application-specific machine vision systems can be identified by attending one or more of the vertical industry-specific trade shows.  These companies will not be found at the more generic machine vision trade shows where one will typically find suppliers of machine vision components of one type or another. Often the vertical industry-specific trade shows will have conference sessions that include presentations on the latest versions of the industry-specific, application-specific machine vision systems.

Application-specific machine vision system suppliers represents one class of suppliers found in the machine vision industry. When one has an application that cannot be addressed by an application-specific machine vision system, then one has to become familiar with the other classes of suppliers so one can distinguish those that are most likely offering products and/or services that can address their application. These classes include:

Vision Processor Vendor–A vision processor is meant to denote a product that is configurable into a machine vision system. It includes products such as general-purpose machine vision systems, intelligent image processing boards or frame grabbers with image processing functionality incorporated, and embedded vision computers/vision sensors and smart cameras. In addition to including frame grabber functionality a vision processor also includes on board computer processing capability. This could be in the form of an embedded microprocessor, FPGA, DSP, array processor, ASIC, proprietary hardware, etc. The graphic user interface is such that little or no reference is made to image processing and analysis. Rather, the interface refers to generic machine vision applications (flaw inspection, gauging, assembly verification, find/locate, OCR, OCV, etc.) and walks the user through an application set-up via menus or icons. These systems may or may not have the ability to get into refining specific algorithms for the more sophisticated user.

A vision processor supplier can use some combinations of:

  • Proprietary frame grabber + proprietary software
  • Commercial frame grabber + proprietary software
  • Proprietary IPBS + proprietary software
  • Commercial IPBS + proprietary software
  • Proprietary hardware + proprietary software

Smart Camera Vendor-A Company that offers a camera with embedded intelligence, such as a microprocessor or DSP, which can be programmed to make the camera behave like a vision processor-based system. The graphic user interface is such that little or no reference is made to image processing and analysis. Rather, the interface refers to generic machine vision applications (flaw inspection, gauging, assembly verification, find/locate, OCR, OCV, etc.) and walks the user through an application set up via menus or icons. Smart Cameras may or may not have the ability to get into refining specific algorithms for the more sophisticated user.

The embedded vision computers/vision sensors are distinguished from smart cameras. A smart camera includes the compute functionality required to execute machine vision algorithms within the camera head itself. The embedded vision computer/vision sensor includes the compute functionality in a separate box with a cameras tethered to the box.

Frame grabber - a board that includes A/D, LUTs, memory to store one or more frames, D/A's, but does not include any on-board processing capability. In some cases, they merely condition the image data out of a camera making it compatible with processing by a personal computer. These boards can operate with either digital or analog cameras.

Image processing board set suppliers (IPBS) - A company offering one or more products, such as a frame grabber with firmware that performs certain image processing algorithmic primitives at real time rates, and off-loading the computer requirements to the firmware from the computer itself. The interface supplied generally requires a familiarity with image processing and analysis, since one will generally start at the algorithm level to develop an application. These boards can operate with either digital or analog cameras.

Machine vision software (MVSW) -a supplier of generic software libraries that can be adapted to many different applications and/or a supplier of software tools for specific applications, for example: OCR, OCV, alignment, robot guidance, BGA inspection, LCD inspection, IC package inspection, component alignment, etc.

Merchant system integrator - A Company that provides a machine vision system with integration services, adapting the vision system to a specific customer’s requirements. A system integrator is project oriented.

A merchant system integrator provides:

1. Turnkey system based on:

  • Commercial frame grabber + proprietary software or commercial software
  • Commercial IPBS + proprietary software of commercial software
  • Commercial GPMV + proprietary software or commercial software

2. Plus value added: application engineering, GUI, material handling, etc.

Significantly, there are many types of merchant system integrators. Not all are dedicated exclusively to machine vision. In fact, most offer machine vision integration capabilities coming at it from a specific set of applications. In other words, they offer machine vision integration as one of many products they integrate. Some such classes of integrators include control system integrators, material handling system integrators, packaging line system integrators (often specializing on a specific industry, pharmaceuticals, for example), symbology reader integrators, robot system integrators, etc.

Generally these type integrators are more comfortable with integrating ‘‘canned’‘ machine vision products offered by one or more of the suppliers of more general-purpose products like smart cameras or intelligent image processing boards or vision processors. In these applications not only is the hardware and software well defined but also the ‘‘staging’‘ – the geometric layout or relationship between the camera and lighting.

Depending on how one defines their requirements one may find one or another type of system integrator as a more appropriate vendor to solicit for the application. Another consideration in the selection of a system integrator would be their experience in the specific industry you are in and with the specific type of application. Still another consideration might be the geographic location of the system integrator. Ideally they should be located close enough so that it would be convenient to monitor the progress of the system over the project cycle.

Should your application not really fit into the ‘‘canned’‘ class then it is best to identify a system integrator that concentrates on machine vision systems exclusively. These integrators have an in depth understanding of the underlying algorithms available either from the hardware supplier or the software supplier and can, therefore, apply them most appropriately and fine tune them for the specific application requirements. They are also very knowledgeable in ‘‘staging’‘ issues and can understand quickly which lighting arrangement will optimize performance, which camera and which set of optics.

When dealing with a system integrator one has to appreciate that it is a team effort. The end-user in-house staff has to supply process knowledge, samples, access to installation site, appropriate skills to take ownership, facilities required, etc. The system integrator provides cost effective labor, technology application expertise, industry familiarity, professional competency, local support, etc. The machine vision vendor provides product knowledge, system expertise and life cycle support. When you look for a system integrator you should look for one that is technology independent. A system integrator should be able to handle machine vision products from a number of different companies so they can select the one most suitable for the specific application.

Questions to ask of a system integrator:
Have you done anything like this before?
What do other clients think of you?
Do you understand my requirements?
Are your skills consistent with my requirements?

In the solicitation or request for proposal/quotation you issue to the companies/system integrators identified and qualified as appropriate vendors with the capability and experience to address your application it is important to ask for a response that demonstrates they have thought through how they propose to address your application. This should include a review of how they propose to address the specification: what are the implications of the variables (appearance, position, etc.) on the ‘‘staging,’‘ image processing and image analysis; how do they estimate the time budget to assure that the throughput requirements can be satisfied; how do they propose to address temperature, vibration and positional error budgets; how do they propose to address the man/machine and line interfaces; diagnostics, calibration, reporting requirements, etc.

In addition the proposals should be requested to address responsibilities of both vendor and customer as well as staff experience and specifically machine vision application engineering skills, project management skills, evidence of good organization and management practices, quality control procedures, system acceptance and performance validation testing, software practices, training materials, documentation, warranty and service policies. In addition relevant references should be sought. It goes without saying that the proposal should be furnished with a schedule and cost.

Having solicited 3 – 4 companies/system integrators to bid on a system for your application, one should systematically review each of the proposals by establishing specific criteria/parameters to use to gauge responsiveness. Given that not all parameters are of equal importance, it makes sense to give each parameter a weight. Those parameters critical to the success of the installation should receive a weight of 10. Others should be given a weight of 1 – 10, depending on importance, with 10 being most important.

Having established the parameters and their weights one can then evaluate each of the proposals against each of the criteria/parameters and assess a score on a scale of 0 –10. By multiplying each score with each weight one arrives at a figure reflecting degree that proposal satisfies the criterion/parameter. By summing up all these values one arrives at a score for the respective proposals. The proposal with the highest score should be the technical winner. Other considerations in the ultimate decision may include cost, schedule, and confidence that company has financial resources to satisfy application requirements and support a system both pre and post installation.

Before making a final decision on a vendor, however, it is advisable to visit at least the two that scored the highest in the preliminary systematic proposal evaluation. This visit will make it possible to confirm impressions of financial viability, technical resources, physical facilities, personnel skills (especially as related to the project – optics, camera, lighting, software, etc.), relevant experience, typical documentation, software documentation practices, quality control practices and the likelihood of meeting the schedule. This meeting should also be an opportunity to mutually clarify issues and gain assurance that the prospective vendor understands the requirements as well as has a familiarity with the surrounding production processes and the industry.

At some point it is advisable to get the names of several references. These should be contacted to assess the quality of their work, ability to meet schedules, support of policies and ultimately would they have made the same decision to use the vendor based on their experience with the vendor. 

So the steps to take to assure success with machine vision especially for an application that is unique or one for which there is not an application-specific or ‘‘canned’‘ solution should include:

  • Become informed on the subject of machine vision
  • Understand production process and variables that may affect machine vision system performance
  • Write comprehensive functional specification
  • Understand nature of different types of machine vision vendors
  • Write RFQ stipulating the kind of details you want covered in a response
  • Solicit appropriate and qualified vendors
  • Use a quantitative approach to evaluating each of the proposals
  • Visit the top two proposers to confirm impressions and assure project understanding
  • Issue purchase order

Using this systematic approach to purchase a machine vision system especially for a new application will yield a successful installation. Where time and corporate resources are lean it may make sense to use an independent consultant. Such consultants can complement staff by writing specification, writing bid package, identifying appropriate vendors, evaluating proposals, evaluating vendors and downstream preparing acceptance test plans.




There are currently no comments for this article.

Leave a Comment:

All fields are required, but only your name and comment will be visible (email addresses are kept confidential). Comments are moderated and will not appear immediately. Please no link dropping, no keywords or domains as names; do not spam, and please do not advertise.

First Name: *
Last Name: *
Your Email: *
Your Comment:
Please check the box below and respond as instructed.

Search AIA:

Browse by Products:

Browse by Company Type: