5 Inspections made Possible with Color Imaging and Deep Learning
Almost 30% of VisionPro Deep Learning applications, Cognex’s PC-based deep learning inspection software, require color imaging. This is because color is an important distinguishing feature for many assembly verification and defect detection applications.
When we launched the In-Sight D900 in the spring of 2020, we knew it was a first-of-its-kind product: an In-Sight smart camera with deep learning technology embedded directly on it. We also knew there were ways we could make the product even better just from talking with customers and taking their feedback seriously.
Whereas the initial models of the In-Sight D900 could only produce monochrome images, the In-Sight D900 Color furthers Cognex’s mission to expand the range and breadth of inline inspections that can now be automated.
Here are five such inspection possibilities and why color imaging is an essential component of the inspection.
Foreign Particle Detection
Picture a peanut packing plant. There is a conveyer of honey roasted salty peanuts being inspected to make sure they meet the company’s standards for getting the privilege of being someone’s Sunday afternoon snack during the football game. The last thing that plant wants is for an unwanted foreign particle to sneak through the inspection phase, which could happen if a small piece of an inspector’s glove breaks off during inspection. That small piece of purple latex glove would be almost impossible to spot in the pile of peanuts with a monochrome imaging vision system.
Assembled kits, for example a medical pouch that contains syringes, bandages, and other supplies, likely contain similar looking objects that are very different in actuality. If a medical kit contains syringes with different medicines such as a flu shot, adrenaline, and more, the syringes themselves could be color coded to indicate their differences. A monochrome vision system would treat these crucial differences as being the same. Only a color image can help further to ensure that the similarly looking but different parts are properly contained within the kit.
Rear Assist Parking Sensor Inspection
Most cars these days come with cameras to help drivers park or go in reverse as part of their standard safety equipment. Those cameras work by having sensors embedded in the bumper and manufacturers need to ensure that those sensors are the same exact color match as the bumper. Many auto manufacturers have 25 or more varieties of sensor colors, including different shades of the same sensor color. It is a challenging inspection for numerous reasons, but one that absolutely requires deep learning and color imaging.
White on white bottle inspection
Whether in consumer-packaged goods or in the food and beverage industry, there’s a chance at some point manufacturers will have to inspect a liquid product that is the same color as the bottle it is contained in. Without the use of a color image an inspection system might not realize the liquid is spilled out because it blends into the container.
Everybody loves pizza. Even frozen pizza. Especially pepperoni pizza. But not everybody loves their pepperoni pizza with stray bits of mushroom on top. When frozen pizza varieties get inspected on the conveyer, a black and white image would make it difficult to determine the correct toppings have been added and especially if unwanted toppings were accidentally added to the wrong pizza. Color imaging, therefore, makes it essential for frozen pizza inspectors to ensure the pepperoni really is pepperoni and not red peppers or onions or salami or any other unwanted toppings.
This is, of course, just the tip of the iceberg when it comes to all the inspections that can benefit from the combination of color imaging and deep learning. If you're curious about getting started with a deep learning project check out our eBook below.