Vision System Design - Innovation Award: Auto-Teaching and Smart Color Segmentation Technology in Blister Inspection

Award Link: https://www.vision-systems.com/factory/article/16737590/vision-system-design-announces-2017-innovators-awards

How our innovation is original and how it fulfills a need in the market that hasn’t been addressed which increases productivity and leverages a novel technology.

Automated Inspection of tablets and capsules packed in blister in not a new requirement, but even today it is extremely challenging task for vision system designers. Since pharmaceutical companies are constantly coming-up with different colors of product (i.e. tablets and capsules) packed in various types of forming foil (i.e. Aluminum, PVC-Clear, PVC-Opaque, etc.)

Examples of few challenging products:

White tablets/capsules packed in White PVC foil
Gray tablets/capsules packed in Aluminum foil
Dark colored tablets/capsules in Aluminum foil
Multi-Color capsules with printing on it.
Transparent Gel capsules in Clear PVC foil.

In addition to this pharm companies today requires high productivity with consistency in quality. So they are investing more and more on high speed machines but when it comes to inspections there is still a lag in productivity and consistency. One of the most important processes in handling vision systems is a teaching process. Teaching a new product in vision system is still a time consuming task and prone to human dependency specifically for bister inspection system. It requires expertise of users on every step, i.e. manually identifying of formed cavities, segmentation of tablet/capsule color, setting optimum tolerances to detect geometric and color defects. Also manual teaching in certain times is a time consuming process and also depends on user proficiency. Even after teaching there are various acceptance tolerances that user has to tune manually to get the optimum results.

After market survey of various blister inspection system available today and understanding pain points of users, our aim was to make this teaching process completely automatic including auto learning of acceptance tolerance. After a constant research and development efforts of 2 years, we developed an algorithm which works for most of the products packed in all types of forming foils. With this today all our customers are very happy with their productively increased drastically. E.g. certain products which took 7-8 hours of users for teaching products earlier, it just takes 2-3 minutes now, even though inspection consistency is maintained.

What are the unique features of this product/system/technology as it compares to existing solutions?

Automatic Teaching for Blister Inspection
- Automatically locating cavities / pocket for all type of foils.
- Automatically identifying colors of product and setting its optimum threshold limits.
- Automatically locating blisters (i.e. group of cavities).
Auto fine-tuning of color segmentation / threshold limits based on multiple images provided by user.
Auto fine-tuning of acceptance tolerance based on multiple good and empty blisters provided by user.
Optional Manual Teaching is also equipped with color segmentation based on smart classifiers i.e. Naive Bayes Classifier.

Even after teaching is completed, user has to constantly keep changing various tolerances during production to keep vision system perform as expected with minimum false rejection. Ultimately this leads to high rejection of blister pack, to destroy this rejection is ultimately a danger for an environment since destruction of chemical waste is hazardous.
Today most of the pharmaceutical companies has started practicing lean process to improve productivity which says reduce waste in every aspect of production process. So manual teaching process which is a waste of time, money and environment has to be eliminated.

Why Auto-Teaching and Auto Tolerance Setting is Innovative and meets the definite need?

Blister pack inspection is all about color based segmentation of tablets and capsules, and then quantifying the segmented region by product geometry and color.