April 04, 2022

Contamination Feature Overview

The team at Pello has been working hard to launch the Contamination Alert system and we are happy to announce that Version 1 is going live shortly. We use a combination of Machine Learning, AI, and advanced algorithms to identify the number one item that our customers say is their biggest contamination issue “plastic bags”. The dreaded plastic bag can create all kinds of problems when it ends up in the cardboard recycling stream from jamming recycling sorting equipment, blocking optical sorters, and generally just being a pain for those folks sorting cardboard and paper from everything else.

Contamination identification has two parts – figuring out if a container has contamination (in this case is plastic bag present in the picture) and then providing a level of confidence in the analysis. Essentially the system outputs an analysis of each image with the decision made and how confident the system is. One of the goals of the Pello team is to provide a high confidence analysis that can be actioned by users. So instead of providing a lot of alerts of which many are not high confidence we want to provide only alerts that we have high confidence of being correct.

How can the Contamination Alert system be used to improve your operations?

  • Reduce contamination events and associated contamination fees charged by vendors by proactively removing contamination items before containers are collected.
  • Identify containers that have high levels of contamination and take action to reduce the contamination by improving signage, training employees and tenants and securing containers to reduce illegal use.
  • Automate your contamination fee billing by using the feature automatically generate and assign contamination events by client or account.

Here are some examples of the system at work:

In this example, no plastic bags have been detected with a confidence level of 98.9%
In this example, a plastic bag (contamination item) has been detected with a confidence level of 96.39%
In this example a plastic bag (contamination item) has been detected with a confidence level of 99.98%

So how is Version 1 performing?

  • based on a test case of 400 images with 200 showing plastic bag contamination and 200 that only show cardboard boxes the system is currently able to identify plastic bags at a 93% accuracy level.
  • The confidence level of the system when it does identify contamination is quite high – on average when a contamination event is identified the system is 88.5% sure that there really is a plastic bag in the container.
  • What happens when a contamination event is identified? Pello can send a real-time alert to multiple users who can take action on the contamination or contamination events that can be stored and analyzed to see which containers have the highest contamination event levels.

Want to improve your waste and recycling operations?