Automatic detection of plastic from construction waste

Important results from the project

The project aimed to test and assess the potential of using a special AI model, developed by Chalmers Industriteknik, to be able to identify and locate hard plastic objects on a conveyor belt in the waste stage. Today, this type of sorting is done manually as there is a lack of good enough techniques and mechanical equipment to trump humans with automated approaches. With promising results, the project could lay an important foundation for continued work and development in order to be able to introduce automated sorting methods in the future.

Expected long term effects

The project is considered to have been very successful in terms of execution and results achieved. The overall result goal of testing and assessing the potential of using the current AI model, built and trained based on prevailing manual sorting of objects in the waste management stage, to be used in this case for an automatic sorting of hard plastic, is considered to have been fulfilled with a good margin. As a preliminary study, the project has exceeded the executors´ expectations and laid a very good foundation for continued work and further development of the technology.

Approach and implementation

Overall, the data was collected over a period of 2.5 months. The data in question is video material recorded with the help of two optical cameras at a waste facility and the manual sorting of hard plastic that is done there. Looking at the promising results and the equipment used for the execution, the project shows that it is possible to carry out high-tech development with small financial resources. This is of course very positive and something that in the long run can lead to more automated process steps in the waste industry and to a relatively modest investment cost.