Choosing the Right Machine Vision Supplier
A Machine Vision White Paper
Teledyne DALSA Posted 04/27/2005
The use of computer vision for test, measurement and inspection tasks has steadily increased with the technology’s ability to perform reliably and predictably in real-life production/ manufacturing environments. As the technology has become easier to deploy and more dependable, OEM’s, system integrators and end users alike are looking to computer or machine vision to help shorten time to market, reduce manufacturing costs, and increase yield. Today’s business environment demands accelerated time to market, shortened product life cycles and heightened responsiveness to customization, all of which represent significant challenges for equipment manufacturers looking to design and build reliable vision systems for truly demanding applications. Faced with these challenges OEM’s do not have the luxury to go through repeated design cycles, or re-inventing machines from the ground up. Rather they must leverage heavily from engineering work that was done previously.
The good news is that some of these challenges can be over come early in the process with the selection of the right machine vision components supplier. This white paper will highlight some of the key attributes you should look for in a machine vision supplier, and why taking the time to carefully assess a supplier’s capability and choosing the right partner will save you time and money in the long run.
THERE’S NO SUBSTITUTE FOR EXPERIENCE
Not all machine vision suppliers are alike. To begin with, it’s important to find a vendor who is focused on machine vision; one with many years of practical experience deploying real systems in real environments. The last thing you want to hear from a vendor is uncertainty or ambiguities when you call with technical questions. A knowledgeable machine vision supplier can help assess your application requirements and specify system components to meet those requirements. For example, look for a vendor who can offer practical advice on the best computer platform to use; on what bus architecture meets your requirements; a vendor who maintains a list of compatible computers in their applications departments and has tested the latest operating systems
A knowledgeable vendor can also offer practical advice on camera selection to meet your inspection requirements and has actively used these cameras in different applications.
Ensure that your prospective vendor is familiar with your market or application from the machine vision point-of-view and look for tell-tales clues that they are credible.
- Are you forced to explain fundamental details to your vendor, or can you speak at a higher level?
- Does the company have white papers/tutorials or editorial coverage on their web site dealing with relevant subject matter that support their knowledge assertions?
- Do they maintain an active network of third party vendors and list these companies on their web sites?
- Are they able to provide a list of third party products that are compatible with their own and back that up with case studies and application notes?
IN IT FOR THE LONG RUN
Is your machine vision vendor in it for the long run? Is there a chance they might leave you high and dry? For example, imagine yourself holding a ‘‘Last Time Buy’‘ notice asking you to purchase the quantity of components that you THINK you are going to need for the next three years? They didn’t even call you; they simply sent the notice over the fax machine, hoping it would reach the appropriate person. Now you are in a position that you have to forecast demand for the next 2 – 3 years, place a volume purchase order with the vendor, take delivery (and, by the way finance the whole thing) and inventory a whole bunch of goods.
While last time buys are a fact of life, the whole situation is managed significantly better by companies who are in the game for the long run. The secret is close communications about product life cycles and availability. How do they handle the product termination process; check their ISO quality procedures about things like last time buys. Are they able to offer viable and efficient migration paths to new generations of products? Do they offer software portability between hardware generations; allowing you to move easily from one hardware product to the next? Check their track record, do they have a history of ‘‘last time buy’‘ notices and finally ask for references from long term customers; were they happy with service and product migrations. Good companies stand on their records.
TRIGGER-TO-IMAGE RELIABILITY
Many non-dedicated machine vision manufacturers of imaging related components do not have the focus, depth, or products to solve challenging machine vision problems.
In many cases, their products will perform poorly in extreme conditions, causing the machine vision system to be unpredictable and inconsistent.
On the other hand, more experienced and dedicated machine vision suppliers understand the long-term issues in and around your applications and engineer their components to perform consistently when the inspection rate increases, or when more precise triggering and measurements are required. Check to ensure the prospective vendor is ISO certified and asked about the specific quality assurance programs they’ve implemented under that program. There is however another important reliability consideration beyond that of individual component performance; it is the reliability of the overall system to perform at a consistently high level of efficiency. System reliability leverages both software and hardware components to track and verify all I/O and internal image processing functions to make sure they performed as expected. System reliability can only be solved through a comprehensive framework that delves into every aspect of the system’s operation from triggering and image acquisition through processing and output.
Consider an inspection application on a high-speed manufacturing line. A part approaches a sensor that triggers the machine vision system to acquire an image.
After some unknown latency, the vision system sends a signal to the camera to begin reading out the image information; more latency, and then the frame grabber begins collecting image data in a memory buffer before passing it to the image processing engine. Hopefully, the part has not moved while the microprocessor was servicing other system resources. Once the image is in processor’s memory, the algorithms convolve the image and a result is issued for action and archival, but how does the operator know that any of this actually happened? What if two parts are placed too close together, or the trigger sensor ‘‘bounces’‘ and sends two ‘‘initiate’‘ signals rather than just one? What if there’s an unusually long delay between trigger and acquisition, or the camera only reads out half of the image?
The normal results are false negatives and high reject rates that cost the end user productivity. To ensure the reliability and robustness of the vision system, there must be a built-in process to track the beginning and end of each event very precisely in time or space, store the tracking data, and react when events transpire out of sequence or contrary to expectations. Only a system that tracks from trigger-to image can guarantee that the data coming into the system is the right data at the right time, otherwise, it’s ‘‘garbage in, garbage out.’‘ Of course it is impossible to design systems that perform at 100% efficiency, so it’s important to ensure your prospective vendor recognizes the importance of system reliability and if so how have they engineered to optimize performance. Are they able to recognize error conditions and is there a recovery mechanism in place to manage and report on those errors?
COMMITMENT TO SOFTWARE
Also consider the supplier’s ability to provide and support imaging software. The core intelligence of any machine vision system comes from its software. Software ultimately controls the hardware performance and drives processing and analysis.
Look for a supplier with an integrated software solution; one that can support new operating systems, and emerging hardware platforms like PCI-X and PCI 64.
Don’t expect much help from a non-dedicated machine vision manufacturer when you want to know how to do sub-pixel measurement, or how to read barcodes on material with changing reflectance. They won’t know and it is often beyond the scope of their capability.
Choose a vendor, who will be there in the future to stand behind their software product, to offer upgrades and fixes; one who invests in software R&D to develop new machine vision algorithms and support new programming paradigms. Check out the vendor’s track record for software; is there a product roadmap and continuous product migration or do they move ad hoc, changing radically from one version to the next? What are their policies and responsiveness to ‘‘bug’‘ fixes and upgrades? Are their licensing agreements clear and concise and do they offer on-line, web-based support?
ESTABLISHED NETWORK OF ALLIED SUPPLIERS
A machine vision solution typically requires many items: lighting, part positioning, cameras, optics and control logic, as well as image acquisition hardware, processing software, and engineering services. Ask your prospective vendor what they do about all these other required components in the system. Be suspicious of any vendor who says they do it all: chances are they don’t. Look for a supplier who understands the limitations of their core competencies but has developed long-standing partnerships with other suppliers to complement their offerings. Find a supplier with an established network of selected integrators, dealers, and representatives to help optimize the performance of these different components.
PERFORMANCE/PRICE AND VALUE DELIVERY
Cost is another important consideration. This would seem easy. You get a quote and go with the low bid, right? Be careful, because getting an estimate of true cost of vision system components is difficult. A low bid often hides the real costs that are revealed only after you’re well into the deployment process.
It’s important to start by setting concise performance objectives for your system.
The big challenge is then matching those objectives with the right technology for the right price. Unless you are very familiar with a product, the product has been designed specifically for your application, or you have extensive in-house system integration experience, you may be in for some surprises. In most cases, deployment issues are more difficult than an inexperienced supplier is letting on. You may end up paying this sort of vendor many times more than their original estimate for services and additional components that were never specified at the start.
A second point mentioned earlier, is that the low-cost vendor is often the one without the product, focus, commitment, and experience to solve your problem. A ‘‘good deal’‘ is not a good deal if it doesn’t meet your requirements. It will end up costing you more to make it right. The reality is, estimating your machine vision system cost is difficult, even with the best intentions of the vendor. Machine vision has made tremendous strides in usability and capability, but it is not a completely ‘‘off the shelf’‘ product for most applications. Development and setup time can vary from a few hours to several man-months for a complex, integrated machine vision system.
COMPREHENSIVE SERVICE AND SUPPORT
Ask your prospective supplier about their service and support programs. Look for clearly defined deliverables. While the exact details of a service and support program may vary from vendor to vendor, here are some general guidelines to look for.
- Inventory reserve for emergency order fulfillment
- Long-term product availability for a technology migration path
- Advanced replacements for immediate failure replacement
- Priority technical support
- Flexible product training-on-site or at vendor’s training center
- Customized hardware and software solutions
- Vision system application engineering consulting
- On-line logistical information for order tracking and account management
- In-depth on-line knowledge base and training tools
STRONG TRACK RECORD
Finally, it is wise to pick a vendor who will be there for the long run. When your machine vision system needs maintenance or you want to upgrade, modify, or rework the system, you don’t want to hear that the vendor is gone. Look for a supplier with a strong financial record, and the long-term customers to back it up. Properly deployed, computer vision can play a key role in making your test, measurement or inspection process more efficient and ultimately improving profitability. Choosing the right machine vision supplier is an important first step toward realizing all the benefits of computer vision technology.
HOW TO LEARN MORE
DALSA Coreco is dedicated to helping customers understand and simplify the complexities of machine vision systems. To that end, DALSA Coreco offers a comprehensive knowledge base of tutorials, application notes and white papers prepared by our own engineers and by third-party machine vision specialists. This knowledge draws from more than 25 years’ experience solving machine vision problems in many domains.
