Vision Rules!

A global brand of personal care products found some of its own plant staff were breaking the rules, letting products with incorrect or misaligned labels slide through the vision system. Sometimes, an operator might be at the end of a shift, or have a quota to fill and be at risk of running out of time. Whatever the reasons, the manufacturer discovered that some plants employed clever, and often low-tech strategies for bypassing the vision inspection process. Sometimes the plant staff would simply shut off the vision sub-system and let the products go through without inspection. Other times the staff would place an image of a good product in front of the camera; naturally, the products "inspected" this way had a 100% pass rate!

 

When this company explained the issue to CI Vision (Aurora, IL), one of their requirements was a mechanism to prevent plant staff from bypassing the vision system. "They needed something fool-proof," explains Tom McLean, COO at CI Vision. McLean calls CI Vision's fool-proof strategy "Anti-Vision Circumvention Technology". It's one of the interesting new features of the company's Pro Series, a line of inspection systems for contract packagers.

Koval's strategy was to make sure that the vision system operated with other automation equipment at all times. In other words, when the labeler is running, the vision system must also be running as well. During operation, the vision system uses image acquisition trigger inputs as a mechanism for determining the product's presence so it can be inspected. If a certain number of seconds pass without a trigger while the labeler's 'run' signal is detected, the vision system faults and outputs a signal to shut down the line. Similarly, when the labeler is off (i.e., there is no 'run' signal) and the vision system receives triggers, the labeler outputs a signal to shut the line down. A fault signal also switches on a red tower light.

 

The most important strategy to thwart the attempts to bypass the vision system is the intelligence that detects if the camera is acquiring fake images. In this case the software checks for movement between the acquired images by comparing translation results. MIL's pattern matching algorithms give the location results of the label's features. Then the locations are compared from image to image over the sequence of acquired images. If a certain number of images indicate no movement between them, the current part and all future parts will fail. The user controls the number of images to compare and the tolerance for movement, and the parts will continue to fail until the movement tolerance meets those parameters.