Food and beverage manufacturers face an array of challenges when it comes to packaging, labeling, and quality assurance. Add high speed lines, and overlooked issues can be quickly replicated across hundreds or thousands of products. From creating unnecessary scrap to making products untraceable or out-of-compliance, these errors have real costs that are hard to prevent with traditional QA methods.
Automated QA vision systems offer a reliable way to ensure every label and package is inspected, even at very high speeds. Able to inspect for multiple issues at once, these systems are much more reliable than manual methods. A popular and versatile inspection is OCR (optical character recognition). OCR inspections can read characters, codes, and even entire words on labeling and packages. These vision systems scan for alphanumeric characters and checks those against a trained database. Common applications include:
- Ingredient/allergen verification
- Correct product label verification
- Misprint or unreadable (untraceable) print identification
- Expiration date verification
- Typo or incorrect printer configurations
But how do you know if your line and products would work with an OCR solution? Not all lines and products are good candidates for high speed OCR inspection. Below are 10 key factors to consider.
Factors for Implementing a Successful OCR Inspection System
Factor One: Line speed
While these inspections can be done for line speeds in excess of 1,000 products-per-minute, there are some limitations and extra equipment required once you get above 500 ppm. Using parallel processing, OCR systems can take as little as 30ms to process an image and make a determination, but there must be room on the line to place your rejector far enough away to allow proper processing time. This means you will also probably need to use part-tracking, as several products will end up between the point of inspection and the rejector. This is typically overcome with a setup that includes high-speed strobe lighting for more accurate imaging, an encoder for part tracking, and a programmable logic controller (PLC) to speed up inspection.
Factor Two: Print Quality
Factor Two: Print QualityPrint quality will greatly affect the ability of OCR systems to successfully inspect products. The higher the variability in print and print quality, the harder it will be to achieve high accuracy thresholds with OCR inspection systems.
The most common side effect of poor print quality is increased false-failure rates, which is to say, good products get rejected off the line because of poor print quality. To combat this problem, you may have to lower the systems accuracy expectation – i.e. how close to perfect does the label need to be to pass? Lowering the threshold to 70-85% correct equaling a pass (good) label will keep false-failure rates under control. However, for highly regulated industries that must ensure labels are correctly printed to a threshold of 90% or greater, this means investment in a high end printer is a must for automated inspections.
Factor Three: Camera Selection
Not all cameras are made equally, and more complex inspections/higher speeds may acquire additional software to make the inspection successful.
Most OCR inspections are done with a smart camera, which typically have their own on-board processing, but these processors have both limitation and throughput limitations. For line rates under 600 ppm or reads under 20 characters, this may be all that you need to successfully inspect products. But higher speeds or more complex inspections will require PC-based solutions for parallel processing.
Factor Four: Software Selection
Another tool for implementing successful OCR inspections is the use of highly versatile softwares instead of a camera alone. These softwares are beneficial because they use algorithms to modify comparison databases and decrease processing time while improving accuracy based on individual aspects of how your printer is performing. For example, the program will start making adjustments based on character-to-character spacing, scale, aspect ratio, and multi-line distance.
Common softwares include FlexOCV (Cognex) and SureDotOCR (Matrox). Some software solutions also now feature an “auto-trainable” option that allows line operators to train the system to new labels/products without the need for an integrator to update the database.
Factor Five: Product Spacing
How closely your products are spaced on the line can be an obstacle to proper inspection. For example, if you what you are trying to inspect is on a surface that is pressed up against other products, reading the labels consistently is not likely to happen. The easiest OCR inspections tend to be one the tops (lids) and bottoms of containers since those rarely are overlapping other products.
Side inspections are certainly doable, but this is where spacing or product presentation to the camera will start to matter more – i.e. could add cost. How challenging this will be goes hand-in-hand with the next factor.
Factor Six: Fixed vs. “Spin-able”
If your products are relatively fixed in orientation – think geometric (boxes, tubs) shapes that don’t spin on the line, most OCR inspections should be very doable. However, slick, round cylinders will have a much harder time getting consistent inspection as they can easily spin and will therefore come down the line at random orientations. In other words, if it’s hard to predict which way the label will be facing when it comes down the line, the more challenging (or expensive) the inspection will be.
Again, top and bottom inspections are the easiest application, because even on randomly oriented lids/bottoms, one camera from above or below the product can capture and read the information. Side inspections with random orientation could require 3-4 cameras in order to catch labels no matter what direction they are facing.
Other Considerations
The success of any vision application depends on a variety of factors. Everything from your plant environment to operator interference can cause vision systems to be successful or quickly get out-of-spec. To learn more about common reasons vision systems fail and how to prevent or fix that, check out this guide on the topic.
The higher your line speed, the more things you want to inspect, and the more accurate/precise your requirements, the more challenging finding an accurate solution will be. This is where an expert integrator like EPIC can help. We can help you balance needs with costs, finding a durable solution with a justifiable ROI that advances your quality inspection.
Depending on the challenges you face, OCR inspections can significantly improve your operations by:
- Eliminating or drastically reduce labeling and date-coding waste by catching mistakes within 10 products or less
- Enabling easier FDA compliance (CFR Part 21) by logging data automatically on every product
- Improving product traceability throughout your distribution network by ensuring codes and labels are readable and traceable
- Ensuring compliance with import/export regulations through verification of correct labeling and dating
- Improving brand reputation and reduce risk of recalls by ensuring all products leave with the correct label
- Improving line speeds by removing QA bottlenecks
If you are ready to discuss OCR solutions, please contact us to talk to a vision expert.
You can also read more about our other automated quality assurance solutions for food & beverage manufacturers here.
Some content sourced from: https://www.vision-systems.com/cameras-accessories/article/16737579/factors-to-consider-for-highspeed-character-reading