Image Pioneer Takes Proofing to a Higher Level
1995; American Bankers Association; Volume: 87; Issue: 3 Linguagem: Inglês
ISSN
0194-5947
Autores Tópico(s)Vehicle License Plate Recognition
ResumoSo far, most banks that use image technology have embraced one facet of it: imaged check statements. These banks capture the images of arriving checks, print up to 18 pictures on a sheet of paper, and send these pages to customers instead of the checks themselves, saving on postage. Although many large banks have implemented image statements, smaller banks have been particularly active, because of the much lower cost of small systems. Among banks with assets of $100-$249 million that responded to ABA's 1993 National Community Bank Operations and Automation Survey, 38% said planned to introduce imaged check statements by 1995. By contrast, few banks actually process their checks using image technology. To handle processing, including proof of deposit, an image system must not only digitize the checks, but try to read the dollar amount typed or written on the check and then encode it without human involvement. Pioneers at imaging the proof-of-deposit function include Barnett, Comerica, Huntington, Signet, Bank of Boston, and Mellon Bank. Mellon's early start Mellon Bank began using an IBM predecessor to current check processing systems in the 1970s, according to David Leach, manager of item processing at Mellon's Pittsburgh site (the only one currently imaging). This 3895 system, which IBM never actually marketed, did not photograph checks and process their images, instead it processed physical checks individually. However, the system did force the bank to redesign workflow to accommodate imaging, and it had software that tried to read the handwritten amounts on the checks. After 15 years of heavy use, the system needed to be replaced. We had a decision to make, says Leach. Would we move forward with imaging or would we take a step back and buy traditional encoding machines? The bank chose to move ahead. In 1986, Mellon began working to develop IBM's ImagePlus High Performance Transaction System (HPTS). The bank and vendor jointly refined the system for seven years. In November, 1993, the bank began actual conversion to the new HPTS system. Three months later, the system began processing checks, slowly taking on greater volumes until November 1994, it took on full check production. Mellon was the first bank in the U.S. to fully use HPTS in production. Were there any difficult moments in the conversion process? Leach won't say specifically, but he does concede that when you're a beta site and you're working with new software, you certainly have more problems than if you were installing something that's already been successfully piloted. Because of its early participation with IBM, Mellon got on-site systems and programming support throughout the process. At any given time, five or six Mellon people worked with the IBM people on the project. Now, the majority of the 250,000 checks that arrive at Mellon's Pittsburgh operations center each night (500,000 on a busy night) are put in an IBM 3890/XP high-speed document processor. A camera--the 3897 image capture system--mounted on this processor photographs the checks and sends those pictures to a 3898 equipped with software that finds and reads the check amount. Reading is fundamental--and tricky Finding the amount is not always easy because check amounts can be all over the place, especially on corporate checks. you can't find you can't read it, points out Leach. Mellon asks corporate customers to put check amounts within a certain area of the check, but they are dealing with hundreds of check vendor printers, and may or may not follow our recommendations, Leach says. There are standards in the industry for check printers, but it's a difficult thing to control. Each customer's going to go out and deal with a check printer, and may be looking for the one that offers the best price as opposed to the one that best follows the guidelines. If it finds the check amount, the character recognition unit then uses algorithms to try to figure out what the handwritten or printed amount is. …
Referência(s)