DECEMBER 2009
| Retail Benchmarking
|
|
Under the Hood
|
|
Product Updates
|
 |
Retail Benchmarking Organized Retail Crime, Bar Code Substitution When I was a child, I had a love of models but lacked the resources to buy them. I'm not exactly proud of it, but on occasion I would go into Toy City, find a model with a lower price and switch the price tags with the more expensive one. I never got caught, but even if I had it was easy enough to explain I'd found the product that way. For obvious reasons, price tags have gone the way of the dodo, replaced with barcodes printed directly on packaging. Barcodes are an improvement for myriad reasons. They are more difficult to forge, remove human error from price entry, and allow prices to be changed without relabeling merchandise. But while they continue to thwart devious children, they are far from fool proof. Today, it is not children saving a few dollars on a toy, but organized groups of criminals printing their own barcode stickers and taking retailers for millions of dollars annually. Barcode substitution is a devious crime because it attacks the part of the system wholly trusted by the cashiers. If the item scans, comes up with a price and a reasonable name, the cashier trusts everything is fine. Many people tout RFID as a superior system for the future, but it, too, has its own security problems. Hackers have been busy making inexpensive devices to read and rewrite RFID tags or even jam them. And the attack is still leveled squarely at the part of the system cashiers take for granted: when it rings, cashiers assume the ring is correct. Vision recognition systems can provide a second tier to validate RFID or barcode product scans using information that is very difficult to forge without altering the whole of the packaging. Our ViPR algorithms actually scan the packaging, just like a biometric device scans a fingerprint. A second, non barcode based or non RFID based validation, is an important tool to reduce front end fraud and is one of the functions performed by LaneHawk InCart. – Justin Beghtol, Senior Software Developer Under the Hood Generic Item Detection LaneHawk has two methods for detecting BOB items. The primary mechanism is to match the item against a reference image using ViPR, our visual pattern recognition algorithms. This works well as long as the reference image is available. There are times, however, when a package changes or when an unusual item is found on the basket bottom and there is no reference image to compare to. In this case. LaneHawk makes use of its generic item detection (GID) algorithms. The relationship between ViPR and GID can be explained with a story of father and son adopting a dog. Suppose that a father goes to the local dog shelter and takes pictures of 100 dogs that he likes. The next day, he sends his son to the shelter and says you can pick out any of the 100 dogs that I liked to be our pet. Just take a picture of the dog, and see if it matches one of the 100 that I chose. The son complies, takes a photo, it matches one of the dogs his father liked, and he brings the sweet pup home. This is how ViPR works. We find items (in this case a dog) that we have seen before. Now for the second case. A few years later, quite happy with the first dog they have, the father decides it is time to get another dog, as a companion to the first. But the father only wants a Poodle or a St. Bernard. So he walks around town over the next few weeks and takes 50 pictures of St. Bernards and 50 pictures of Poodles. Then, he sends his son back to the shelter and says, choose any dog, and take its picture. If the dog in the shelter matches as a Poodle or a St. Bernard, then we can get it. This is how GID works. The algorithms are looking for something that is "like" what they have seen before, but not necessarily "exactly the same thing". For GID, the computer uses a fuzzy logic comparison. Let's take this out of the dog world and into the packaged good world. In GID recognitions, there are two parts to the fuzzy comparison. The first is what we call the appearance vector. This is an array of numbers that is a measure of what the image looks like in general. You can think of the appearance vector as a blurred out version of the image. The second part is the motion between successive images. And to finish off a GID recognition, the appearance vector and motion are fed to a pattern classifier which decides if a BOB item is present or not. See, more than you thought you ever wanted to know about visual pattern recognition. – Bob Boman, Senior Software Developer Product Updates LaneHawk InCart LaneHawk InCart is a new member of the LaneHawk family product line. InCart is designed to do from above what LaneHawk BOB does from below — recognize the items going through your check-out lane and notify the cashier in real time when the item has not yet been paid for. InCart uses the same patented ViPR recognition engine as LaneHawk BOB, and so can run on the same server to capture items left on the top basket of a shopping cart, in parallel with LaneHawk BOB detecting items on the bottom of the basket. When InCart recognizes an item, it first checks to see whether the item has already been rung up. If, by the end of the transaction, the item has not been scanned, InCart notifies the cashier that the item was missed. And, like LaneHawk BOB, InCart can produce several different reports to track cashier usage in order to target suspicious activities and reinforce best practices. Because LaneHawk InCart is looking at a larger region from an overhead camera, and is examining all items present in the basket, it has many more items to look through and recognize. For this reason, the ideal environment for LaneHawk InCart is retailers with larger products and lower SKU counts, or who desire to target a small segment of their SKU's to focus on high ticket and/or high theft items. Some examples include home improvement, electronic merchandise, and warehouse supermarkets. – Dr. Jim Ostrowski, VP, Product Development LaneHawk, A Brief History On January 13, 2006, Evolution Robotics Retail, Inc. was formed to bring object recognition-based solutions to the retail industry. Our first product to market, LaneHawk™, is a revolutionary bottom-of-basket loss prevention solution based on Evolution Robotics' award-winning visual pattern recognition (ViPR) technology. LaneHawk™ uses ViPR technology to recognize grocery items by analyzing the printed patterns on their box, instead of using the barcode. ERR's ViPR technology is the same software that is being utilized by the US military for threat detection and by Asian robotics companies for navigation. LaneHawk today is deployed in more than 8,000 lanes. Retailers across the United States are utilizing the latest technology to combat shrink. As evidenced by our growing customer base, LaneHawk is fast becoming the solution of choice. Additional products, which include our ViPR algorithms, include LaneHawk InCart (for catching sweathearting and items left on the top of the basket), ShelfHawk (for identifying OOS and shelfsweeps) and TunnelHawk (an improved front end self checkout system.) – Colleen Lindsey, Manager, Marketing |
|
 |
© 2007-2010 Evolution Robotics Retail, Inc. All rights reserved. The Evolution Robotics Retail logo is a trademark of Evolution Robotics Retail, Inc. All other trademarks are the property of their respective owners. Privacy Policy.
Evolution Robotics Retail, Inc. | 433 N. Fair Oaks Avenue | Pasadena, CA 91103