Machine Vision in Document Reconciliation Applications

Remember when you received a piece of junk mail and it was addressed to someone else at your address – someone who had not previously lived at your address. Today that does not happen as often and we can attribute that to the use of machine vision in mass mailings. Machine vision is integrated into automated laser printers and mail insertion systems. The systems typically read, match and verify human readable alphanumerics and 1D and 2D bar codes to assure proper mailings.

Systems can generally find the address space, reconcile the addressee with the address and make sure of any sequential numbers associated with the mailing. They can also assure that the envelope and insert are addressed to the same person at the same address.

These systems operate at high speeds - tens of thousands of pieces per hour. Typical installations use multiple camera arrangements. Some of these systems can operate on UV or IR data meant ultimately to be invisible, essentially reducing any real estate that might otherwise be required to achieve the equivalent reconciliation with comparable visible data.

Major direct mail houses, credit card companies and other companies engaged in mass mailings have embraced these systems. When an error is detected, further printing is stopped leaving it up to the operator to correct the problem resulting in the elimination of scrap and rework.

As with other articles in this series of articles on somewhat unique but substantially replicated systems, we asked known suppliers of such systems to provide answers to questions related to this application. The following were kind enough to provide answers.

Guillaume Bernard – US Sales Manager – Axode

William T. Riley – Vice President, Sales and Business Development – Formscan

1. How would you describe your machine vision-based document/mailing integrity verification systems? Specific products you offer that address these applications?

[Guillaume Bernard – Axode] Axode provides a full range of products for the mailing and printing market from low cost mono camera based system like the D:cod or the Axode 60 to more advanced solutions like the 1080 system which can handle up to 24 cameras (with the capability of mixing line-scan and matrix cameras into the same unit). The mid range product is defined by the Axode 62 (up to 2 matrix cameras) or the 180 system (up to 8 cameras), both systems can handle various applications such as: read and print, selective feeding, sequence verification, output scanning, matching, etc.

In addition to this vision system, Axode also provides a PC based unit for Real Time Database management, production reports and automated document factory (ADF) management.

[William T. Riley – Formscan] Formscan has been in the field of applying the benefits of machine vision technology specifically to the document processing market exclusively for over 30 years. We have extensive worldwide experience in both the end user and OEM fields. By this I mean that we have tremendous hands-on expertise applying machine vision technology to specific end user document production applications such as printing, finishing and mailing in both in-house and large industrial sites, and we have also packaged the solution systems into OEM suitable kits to be added to their base platform.

Formscan recently introduced ‘‘MagVision’‘, which allows MICR images to be read within envelopes or behind other ‘‘image blocking’‘ pages. This will enable use of cameras within closed face match mail applications, read and print, and signature to cover collation.

2. What specifically differentiates your machine vision products that address document/mailing integrity verification?

[Bill] Formscan has excellent experience in meeting and satisfying the needs of many unique document applications, and the technical expertise to package technology for wider distribution through an OEM. We enjoy long standing partnerships with the most recognizable players in the document production field such as Xerox, IBM, Kodak, Pitney Bowes, Bowe, etc. We support many installations first hand including installation, training and documentation and provide on going support. And because the system is Windows based and designed to be more easily adapted and supported, we also provide field level support training to our OEM’s who then provide first level support, coupled with our second level support.

[Guillaume] All our systems are developed and manufactured by Axode. This leads us to have a true real time processing of the image. The result is one of the fastest and most reliable/stable reading devices in the market. This specific approach (of a true manufacturer and not a component assembler) is particularly appreciated by our customers (end users and OEM) since it allows us to custom design our solutions.

3. What are the specific document/mailing integrity verification applications your machine vision-based products address?

[Guillaume] Each of our multi-camera based systems comes with a complete software package that includes:

• Matching of multiple pieces either by reading and ID # (OCR, Barcode, 2D Barcodes) or with the UNIQUE Address Block Matching.
• Read and Print capabilities
• Sequence verification
• Print quality control (address block for the mailing/finishing environment of test pattern verification for the high speed printers)
• Output scanning (by comparison with an existing database or by collecting the read data into a database)
• Postnet verification (Axode has designed a low cost solution that verifies the print quality of Postnet Barcode by comparison with the Merlin specifications).
• Scratch off label positioning and integrity (for the gift card, loyalty card, prepaid phone card market).
• Selective feeding
• Accumulation/End of Group Management
• Mark detection/Postal Break detection for sortation

[Bill] We typically integrate commercially available cameras with our own decoding engines and algorithms, hosted on a Windows PC. Consequently, installing, training, ease of use, ease of adoption are quite simple. Likewise, our material supply and support has been simplified by evolving from custom circuit boards, illumination, cables and industrial computers to more standardized commercially available components; our intellectual property is in our application of the technology and our software that makes it all happen efficiently and economically. We have worked extensively with a wide variety of sensors, scanners area scan and line scan cameras, and we enjoy a particular familiarity and excellent field adaptation of line scan technology. Line scan gives the client a more economical and efficient solution especially where one line scan camera can replace multiple area scan cameras. Having said that, we provide the appropriate solution for every application, be it sensor, scanner, area or line scan based. Different applications, be they narrow FOV inserter or finishing document machines or wider FOV print or collating applications are simply better suited to different scanners and cameras.

4. Can you briefly review the underlying principles associated with your machine vision-based products? Specific hardware arrangement and software?

[Bill] Formscan systems are built on Windows technology, the most popular and robust operating platform for business. Yet familiarity with Windows routines is not a requirement for operation or operators – Windows simply fits better within the business IT environment. All of our setup routines are simple mouse clicks from a default program, and this job setup is done once per job within minutes and then saved for future runs. Similar jobs require even less setup as they can be versions of other existing jobs, yet setup and saved uniquely. With commercially available machine vision components coupled by Windows technology driven by our own software based on decades of use and improvement the setup and performance principles are vastly simplified for the end user.

[Guillaume] All of our systems are driven by our own real time hardware and a custom Linux based software. As with any industrial machine vision system or controller the operating system is transparent to the customer who sets up the system through a touch screen interface. Display of the image and operator interface are included into the same interface.

5. What are the critical performance parameters of a machine vision-based document/mailing integrity verification system that a customer should understand? Throughput? Performance in terms of false rejects or accepts? No reads? Mis-reads? Other parameters?

[Guillaume] Due to the real time processing, our systems can read up to 360,000 documents/hour. It takes about 8 to 10 ms to read and transmit a 12 digit OCR number. In terms of misreads ‘‘created’‘ by the system we consider that under perfect condition, we obtain a false rate of 1 to 100,000 reading.

[Bill] In terms of setting customer expectations and what can be done to ensure better results, it’s probably a very familiar refrain: the human eye is marvelously intelligent, yet a camera can only see and process what it is designed to see and process. Beyond this, it is critical to listen and understand what the specific application requirement is, and then document that understanding back to the client within a statement of work. We believe we have the technology to address 90%+ of known document processing applications – and the remaining 10% can be adjusted and provided to meet specific, unique needs. All clients process document work in very similar fashions, yet each is different unto themselves. So we continue to expand and grow our technology while maintaining and efficiently adapting to each unique production process – usually driven by unique requirements for different customers. No two production lines seem to be the same, so we recognize that up front and detail the specific customer’s requirements, structured as objective standards. And then the machine vision technology either does the job, or not.

6. Are there specific application issues that a customer should understand to optimize system performance? Font styles? Paper properties? Gloss issues? Paper color issues? Ink color issues? Type printer issues (laser, ink jet, etc.)?

[Bill] Use OCR fonts; use generous spacing and clear areas; allow the camera to literally see the image clearly; provide a well-printed image.

[Guillaume] As with any vision system we do have limitations in terms of font size, styles, etc. Usually we work with our customers to define these specifications beforehand by providing them with a detailed print specification document but also by testing their live samples.

Our 15 years experience has lead us to come up with creative/flexible solutions for reading various ink color (yellow, red, green, black) on various substrate/background and also through cellophane. Also we have designed a full package of filtering tools that help our customer get rid of noisy information such satellite dots (for inkjet printers), background information, etc.

7. Are there specific hardware parameters that a customer should understand? Optics issues? Camera issues? Lighting issues? Software issues? Etc.?

[Guillaume] One main issue in the mailing industry is sometimes related to the design of the machine we are integrating our systems into.

[Bill] Don’t confuse an intelligent camera system with your home video camera that only sees but does not interpret images. Most importantly within document applications is: know at all times that the camera is processing a unique VARIABLE image. It may contain like characters as within a sequential number, yet to the camera each new image is fresh, different and unique – and that is how they process the image – one at a time, each and every time, each and every document.

8. What are the skills required to set up a job and operate a machine vision-based document/mailing integrity verification system?

[Bill] Operator training is easily done during the installation process, including the setup of typical customer applications. We have continued to refine an extremely efficient and simple setup interface, so that even a basic understanding of English is tolerable. Password protection is included to prevent accidental and unauthorized access to the setup systems, and this benefits from supervisor type skills. Yet the setup and daily operation of the system is designed to be easy and comfortable for today’s document factory operator.

[Guillaume] Thanks to the user-friendly interface it takes less than 10 minutes to set up a new application (from the positioning of the cameras and lighting to the set up of the parameters) for a regular operator. There is no need for this operator to have any knowledge of a Windows/Microsoft based computer!

9. How do you support your products – training, documentation, warranty, post installation service, software revisions? Are these free or is there a fee?

[Guillaume] Once installed at an end user site we usually conduct training that can range from 1 hour (for a straight application like sequence verification or matching) to 4-6 hours for a more advanced application (like print quality control and monitoring at the output of a high speed printer).

An operator manual plus a service report and a customized application note (related to the application run by the specific software) are provided with any system. Warranty on all systems/parts is 12 months. There is no real need for upgrade or maintenance contract.

[Bill] Training, system documentation and application documentation is included with every system, and warranty engages 30 days after installation. We do offer ongoing support packages, more to take advantage of regular and constant software upgrades, as well as for follow on training for new operators and applications.

10. Are there market changes that are driving the adoption of machine vision-based integrity verification systems?

[Bill] We see the expansion of public privacy concerns, the passage of health and financial compliancy legislation such as Hippa, Sorbannes Oxley and Graham Leach Bliley as setting new mandates for the adoption of machine vision to create reliable audit trails and accountability as market drivers. Higher service expectations from our clients’ customers who want and need to know who, what, where, when and exactly how much, is also driving camera integration into document processing applications. Software packages that claim to do the same job as cameras are pretty good, but they are not 100% reliable in a 100% world; for example, legal document applications specifically require the use of cameras that audit every single document, and provide that piece level accountability.

[Guillaume] The increasing need for security and tracking is actually driving the decision makers of the mailing and printing industry toward more elaborated verification/QC systems. One of the answers to this achievement is actually based on vision systems. This is true for both the printing industry (where higher speed, use of color and personalization of the piece requires more verification) but also for the mailing industry (where the message has to be carried in a proper way and addressed to the right person).

11. Are there market changes that will require changes in the machine vision technology? What impact do these market changes have on the technical requirements (specifications) for the machine vision technology? And how will these machine vision systems have to change to address these more demanding requirements?

[Guillaume] The increased demand for high-speed personalization but also the use of color leads the various providers to address the issues that can be related to these technologies. This is why we have been actively working with inkjet and laser printer manufacturers to understand and address their specific needs. One of our main goals is to be able to compare the printed information with what was supposed to be printed by comparison with the data sent by the printer controller.

[Bill] Certainly recent compliancy and privacy federal legislation, and the growth of ADF systems wherein every page of every document needs to be identified, tracked and audited have created new and expanding opportunities for applying machine vision to document production workflows. Since we use commercially available components, Formscan can focus our talent and experience towards application development for these new challenges. We leverage the regular technical improvements from PC performance, improvements in Windows software, improvements in camera and lighting technology with our own application experience and development to anticipate change, and meet and exceed our customers’ expectations.

12. As a supplier of machine vision-based document/mailing integrity verification systems what are some challenges you face in marketing these machine vision systems?

[Bill] Machine vision systems deliver measurable value at many levels to a client – understanding what is most important to your client always helps to advance the marketing effort. More and more, with competitive service demands and compliancy and audit mandates, customers need to account for the real time production status across the multi-step document process. Here is where our Sentinel ADF solution delivers even more compelling and measurable value to clients. Listen. Understand. Document the statement of work. Deliver measurable results – and provide a modular and scalable growth path.

[Guillaume] It is always a challenge to get the customer to understand that a vision system is here to improve his production not only in terms of quality but also throughput.

13. What advice would you give to a company investigating the purchase of a machine vision-based document/mailing integrity verification system?

[Guillaume] The prime advice would be to clearly specify the existing application but also the future needs by first discussing with its clients and then approach one of the vendors who are specialized in this industry. Once they have found one or more vendors they should be able to work very close with them by sending test samples, and get all kind of recommendations as far as integration, reading specs, etc.

[Bill] We advise new customers and existing customers with new applications to set their expectations clearly, and realistically. If the intent of a machine vision system is to see, decode, log and audit every single document within a print, finish or mail line, let it do that without encumbering the system with too many additional requirements. While expanding the extra machine vision application is likely to be beneficial, walk before you run, and let the system establish its primary value. Networked machine vision systems within workflow solutions such as Automated Document Factories are a growing and needed expansion of camera technology, and hold the promise of great return on investments for compliancy, service and process improvement. And like the more simple machine specific applications, clients ought to clearly understand what they want from the ADF solution and document that in a statement of work. Then this SOW can be staged in phases that deliver clear value for every step in the system integration. New technology promises even easier and wider spread adoption of machine vision technology for document processing.