New Technology Brings New Solutions for End Tallying
End tallying bundles is a pain point for most hardwood businesses. It’s critical to account for every inch of inventory, but getting accurate measurements on random width/random length bundles is typically a manual process. It can be time-consuming and has the potential for human error. There has always been an opportunity for technology to solve this problem. The latest end tally solutions will leverage object recognition technology.
Object recognition is a way for computers to interpret images. It’s advanced technology, but the concept is simple: teach software to identify objects based on shape. A good analogy for how this works is the way people recognize road signs. When you see a red octagon, you know it means “STOP” without ever reading the words on the sign. Our brains instantly identify the item based on color and shape alone.
Object recognition technology is already in wide use for consumer products. Facial recognition is a good example. Ever noticed how your smart phone can identify people in the photos you’ve taken? The software in your phone identifies human faces by looking for a specific cluster of shapes (eyes, the mouth, nose, forehead, etc.) Then, it catalogs each face based on “landmarks” such as the shape of someone’s jaw, the relative position of their nose, even the distance between their eyes. If the same combination of landmarks appears in multiple photos, the software determines the photos are showing the same person and categorizes them accordingly.
The reason object recognition software is so effective is its use of machine learning. As programs are fed more information, they become more accurate. Let’s say you want to teach your software to pick out photos of happy people. You’d start by feeding the program images of smiling faces and telling it “these are happy people.” If the software showed you photos of people frowning, you’d simply tell it “no, those aren’t happy people,” and feed it more photos of smiling people. Eventually, the software would figure out happy people have mouths that curve up (smiling) instead of down (frowning).
So, what does any of this have to do with hardwood lumber?
Object recognition technology has huge potential as an inventory measurement tool for hardwood businesses. It would be possible to develop apps that could recognize the widths of individual boards in a bundle of lumber – all from a picture taken by a phone. The people who end tally would use a mobile device to take a photo of the bundle, and the app would determine the exact or rounded width of each board in a matter of seconds. No specialized equipment required, just a phone or tablet.
Hardwood lumber does present a unique challenge for object recognition software because every board is unique. Irregularities may initially confuse some programs. This is why the software’s ability to learn is so valuable.
Consider the issue of splits. An object recognition program might initially interpret splits as breaks, and count one board as two separate pieces. When you correct the program, (“no, that’s a single board”), not only would the app correct that record, it would remember your correction when interpreting similar images in the future. As the app is used more frequently, the more information it would gain and the more accurate it would become. Eventually, it would learn that if a split runs diagonally or if there’s a matching woodgrain on either side of a vertical split, the image is probably of a single board.
In the coming years, the hardwood industry will see more object-recognition software products enter the market. Businesses interested in improving their end-tally processes should keep an eye out for solutions that fit their operations.
This article originally appeared in the June 2018 issue of Hardwood Matters magazine.