By David Tillett
Omron offers a wide variety of vision systems, from bar-code readers to simple sensors to smart cameras paired with a lighting system and inspection software, to systems that employ Omron’s advanced AI algorithms.
Felix Klebe, marketing manager for sensors and advanced sensing with Omron Automation Americas, said that while many companies offer systems that have some level of AI capability, the ease of use of AI is all over the map. Some AI systems require the availability of a very knowledgeable software engineer or vision specialist, but Klebe said Omron’s AI is intended to be very easy for someone at a technician level to implement.
“So that’s … our positioning on AI, is that you’ve got to be careful not to over-engineer it and make it so it’s very difficult to implement,” Klebe said. “It’s sort of ‘OK, how much do you want to understand — or be forced to understand — what’s going on behind the scenes?’ ”
Klebe highlighted the company’s FH Series vision system, which launched last fall and offers two levels of AI.
The standard version of Omron’s FH vision system is Fine Matching, which can learn to discern a good product from a defective product after being shown a pool of between 100 and 200 images. It uses an AI-based image filter that can avoid overdetection, which is when a vision system focuses on every detected variation from a known good sample, including details that are irrelevant for identifying a defective product. This can lead to products errantly being judged as defective. Using images of good products with variations, the AI in Omron’s Fine Matching tool can learn to only focus on defects, such as the presence of foreign material, and exclude irrelevant details, such as minor positional differences or light variations.
The second level of AI is the Scratch Detection filter, which users can add by paying a one-time software license fee. Unlike the Fine Matching tool, there is no need for users to train the Scratch Detection tool — it uses a pre-developed AI model to identify scratches and blemishes that are difficult to detect, especially for textured surfaces or in cases where the lighting isn’t providing good contrast. The tool extracts and highlights the pixels it considers to be a defect using an AI-based binary filter.
“So, we have sort of this two-step approach, which lets you start small,” Klebe said. “If you’re trying to basically just improve your standard vision application, implement a little bit of AI to minimize … those false negatives or false rejects. And then if you want to get a little bit more into it, we have this next level up, which is the fine scratch detection.”
He said with scratch detection, the system has more control over the lighting system, adjusting lighting intensity and colors as needed depending on the product surface material and color to reveal scratches that would be difficult to detect with the human eye. “So, those unseen defects or difficult-to-see defects, that can be quite important,” Klebe said, pointing to products where the scratches could cause a safety issue, or for higher-value products where any defect could prove costly.
Catching problems earlier
And while performing a final inspection of a finished product is important, Klebe said that being able to perform inspections at earlier stages in production can be valuable to manufacturers.
“And being able to catch that very early in the production and not after the phones are already in the consumer’s hand or at a wholesaler or retailer or something of that nature,” Klebe said.
“So that’s sort of the story on how we can use the AI,” Klebe said. “There is the advantage of using a well-proven vision system so we’re not coming up with something that’s sort of on its own and brand new that has no installed base. We’re able to basically take existing hardware and software that maybe is already being used by some of our customers, add this new software revision, and put them in the business of using AI. So, the implementation costs are very, very reasonable compared to starting brand new with new hardware and total new software system.”
Klebe said Omron’s FH system is adaptable. Within the last month Omron launched software that basically builds onto its FH system to enable 3-D robotic bin-picking applications. He said the software includes a wizard that can import CAD data for the part, and then communicates between the FH vision software and the software the robot is running.
He said that by mounting a lightweight FH-SMD series 3-D camera directly to the collaborative robot, the FH system can quickly identify individual parts in a bin. He said Omron’s technology is very fast compared to some competitors’ products and that mounting the camera directly to the cobot instead of using a traditional overhead gantry mount offers advantages, including improving visibility within the corners of a bin.
Currently, the software supports Omron and Fanuc robots, but he said new iterations are in the works to support more manufacturers’ robots and bring the ability to sense smaller objects.
Klebe gave the example of a bin of injection molded parts, from which the cobot could pick parts to put them into a package or onto a conveyor.
He said the system could potentially be used for more complex tasks, such as assembling kits with multiple different parts. As an example, the cobot could draw one part from bin A and place it in a tray, draw a different part from bin B and add it to its location in the tray, and so on — depending on the size of the bins and the robot’s range of motion. He said that a lack of range of motion could potentially be overcome by combining the cobot and 3-D sensors with a mobile base.
Coming into focus
“So, then the other trend, of course, is higher-resolution sensors,” Klebe said. “So, we’re always improving and trying to be able to inspect smaller and smaller items, or larger fields of views or longer distances.”
Klebe said the FH system is a PC-based system that can support sensors with resolutions up to 20 megapixels. It also can be used with up to eight cameras, which provides great flexibility. The FH system could work with cameras deployed in many different combinations — so multiple cameras could be used to capture a single object of interest from different sides, or each of the eight could be focused on different objects from a separate production line.
Klebe said that processing speed becomes a factor in applications where products are moving quickly, such as in food or beverage lines, where a smaller, more compact system may offer an advantage.
He introduced “The baby brother of the FH, which is called the FHV7, which is an all-in-one smart camera with the lighting, the lens, the sensor, all integrated in one package, but then it uses the exact same software as the FH. So, you could have let’s say, a single software application that’s been written and then applied on the high-end FH system or more the mid-range FHV7 system.”
The FHV7 can toggle between different colors of lighting, which allows it to easily provide the best contrast to deal with different colors and materials as well as the lighting that can be found in a mixed production environment.
“Moving down the range a little bit … we have a lower-range smart camera, which we call MicroHAWK,” he said The MicroHAWK came to the lineup through Omron’s 2017 acquisition of Microscan. Klebe described the MicroHAWK as an entry-level smart camera, but said it offers integrated lighting, liquid lens autofocus technology, and a compact design for tight spaces.
Klebe said the FH system can handle resolutions up to 20 megapixels, the FHV7 system offers up to 12 megapixels and the MicroHAWK tops out at 5 megapixels.
Tackling multiple tasks
Klebe also emphasized the ability of smart cameras to conduct vision tasks as well as bar-code reading and optical character recognition. This allows them to confirm that text is readable, formatted correctly (especially important for medical and pharmaceutical applications) and check that the bar-code information matches the human-readable text.
Klebe said combining these capabilities in a single tool offers many advantages. “You’re saving the cost of the hardware, the physical space, and you may be checking at an earlier point in the production so that it can be flagged,” he said. “The further downstream you go, the more cost there is to resolve the issue.”
These capabilities also allow for traceability. “The whole point of a system like this is to minimize the impact of recall … if we look at the causes of recalls in many industries … labeling is often in the top three or four most common reasons for recall — mislabeled product or missing information,” Klebe said.
And these capabilities can be integrated into other equipment. Omron’s V275 Series of print quality inspection system can be used with a laser marker or a thermal printer to make sure what’s being printed meets all required metrics. For example, Zebra-brand desktop printers are widely used by manufacturers for label printing. The V275 can be integrated directly with the printer and check all quality metrics, and if there is a problem, it can stop the printer and alert the operator that there is a problem.
“That process of using a vision system to replace or augment historical random inspection — now we’re doing 100 percent inspection,” Klebe said. “So, in the past, maybe they inspected the beginning of the run and the end of the run, but there in between were a couple thousand items where it wasn’t inspected, and the quality of that printing, or that marking does vary throughout a run.”
Klebe hinted that soon Omron will offer higher-resolution code readers that will be able to take a single image that will capture hundreds of bar codes of products that are within the field of view. He said marking technology has advanced and codes can be printed on parts that were once too small to be coded. Rather than take individual images for each of 200 printed circuit boards on a tray, for example, Omron’s new technology will allow all of the barcodes to be captured in a single image, decoded quickly and used for that entire batch of products.
David Tillett, associate editor
Omron Automation Americas, Hoffmann Estates, Ill., 847-843-7900, automation.omron.com
