AndPlus acquired by expert technology adviser and managed service provider, Ensono. Read the full announcement

How Custom Machine Vision Made Banking More Convenient

Apr 2, 2018 9:05:00 AM

shutterstock_1011978604 smIf you’ve been in a bank lately, you’ve probably noticed there aren’t many tellers—perhaps two or three at any given time, tops. Many banks now encourage their customers to use ATMs instead for most of their banking needs. (Don’t weep for the bank tellers, though -- because of the phenomenal growth in the number of bank branches, there are actually more bank tellers employed in the U.S. than ever before.)

ATMs, of course, have been a convenient way to obtain cash from your bank account for a long time. Recently, however, they have become a convenient way to get money into your account—thanks to the development of custom machine vision algorithms.

ATM Deposits: A Brief History

Until machine vision algorithms came along, making a deposit at an ATM was a bit of a joke. You would put your currency and checks in an envelope, write your name and account number on the outside, and put the whole thing in a slot when the ATM directed you to do so. A human would still have to manually open each envelope, transcribe the account number into a computer terminal, count the money, and process the checks, just like an ordinary visit to a bank teller. Your deposit transaction would be pending until all of this happened—the next business day, maybe. So, if you made a deposit after hours on the Friday of a holiday weekend, you wouldn’t have access to your money until Tuesday. And woe be unto you if you wrote your account number wrong on the envelope.

It wasn’t much more convenient than going inside the bank, standing in line, and making your deposit the old-fashioned way.

Custom Machine Vision Algorithms to the Rescue

Happily, advances in imaging hardware and machine vision algorithms have made all of that a distant memory. A fast-increasing number of ATMs can now count your deposited cash and credit your account right away. Although this is pretty cool, it’s not all that remarkable, considering there is not much variation in U.S. currency; $20 bills all look pretty much alike.

What is remarkable is the fact that these ATMs can read checks and determine the amount. First, there’s the issue of locating the amount on the check, given the endless variety of paper check sizes, designs, and layouts. Moreover, some people still hand-write checks, and some of them have less than stellar penmanship. But most ATMs can accurately interpret even the wobbliest handwriting. They do make mistakes from time to time, but the error rate is quite low, and most deposit-taking ATMs allow you to review the image and verify the amount is correct.

How does it work? After the check is imaged and the algorithm locates the amount, it looks for the boundaries of each character, and analyzes each one, matching it with character shapes it knows about through machine learning or some other technique. When all the characters are identified, the algorithm spits out the amount.

In the case of checks, it helps that there are some rules about how the amounts are written. For example, in the U.S., the numeric amount of the check is written in dollars and cents, with a decimal point in-between, so it’s a safe bet that the two right-most characters are the cents and the rest are the dollars. These rules help simplify the algorithm to some degree.

A similar thing happens when you use the check deposit feature of your bank’s mobile app. Snap a picture, and the algorithm goes to work. It’s a bit more challenging in this case, because of the varying image quality coming from different devices, but the apps still do a pretty good job of figuring out the amount.

Just One Example

This is just one example of custom machine vision algorithms making our lives easier and more convenient. Similar technology enables self-checkout terminals at grocery and department stores, informs Facebook’s facial recognition features, and will soon help self-driving cars navigate through traffic. When devices can “see,” and understand what they’re seeing, it opens up whole new worlds of possibility.

Abdul Dremali

Written by Abdul Dremali

Abdul Dremali is a key content author at AndPlus and a driving force in AndPlus marketing. He was also instrumental in creating the AndPlus Innovation Lab which paved the way for the company’s leadership in Artificial Intelligence, Machine Learning, and Augmented Reality application development.

Get in touch