BottleID

BottleID

Video
Team name / Company name: ProstěTeam
Team leader: Vojtěch Jedlička
Challenge: no. 2: Bottle deposit scheme
Problem: Current RVMs allow people to return uncrumpled PET bottles and cans. Crumpled bottles are a problem, because the machine reads the EAN code, which is unreadable after crumpling. And most of the people here in Czech republic do crush it. So the problem is how to make RVMs accept crumpled bottles and cans, while keeping it economically reasonable, easy to implement and fraud resistant.
Solution: Our solution is to print a short code containing the required information about a bottle/can on the bottom of the bottle which doesn’t usually get crumpled. The code is printed (using the same technology as the expiration date) multiple times to ensure it gets read successfully. The RVM reads this code using machine learning algorithms and pierces the bottle, which makes the code unreadable.
Impact: People will not have to carry uncrumpled bottles, which will lead to greater involvement in the recycling of returnable bottles. According to statistics, over 90% of bottles are returned (if a deposit is provided), but they usually have to be uncrumpled and whole. This technology - allowing returns of crumpled bottles could further increase return rates and facilitate recycling for those who consider carrying uncrumpled bottles a problem.
Feasibility: Our solution doesn’t require major changes to the production lines nor the RVM machines, thus being relatively cheap to implement. The only required thing is a machine learning algorithm to read the code on the bottom of the bottle. Computer vision is commonly used in other industries and it has proven itself to be reliable.
What you built: We built a prototype of a system to reliably read info from deposited bottles/cans in any shape, made a video and a presentation of the concept.
What you had before: We started from nothing, we chose this challenge on the spot. We received guidance from the Mattoni mentor team and the documents they provided to everyone. Thanks to PrusaLab for helping us with the prototypes.
What comes next: We hope to improve the fool-proofness of our system and to tune minor details.