Ing. Matěj Hužvár

Efficiently sorting individual materials from heterogeneous construction debris poses a persistent challenge in realizing the vision of a circular economy. This paper introduces a technology employing RGB cameras and mass sensors operating in 3D, for rapid and accurate classification of materials. Data is processed through specialized software leveraging extraction techniques, machine learning, and computer vision algorithms. Our technology achieves material determination accuracy ranging from 90% to 96%.

The efficient classification of materials in construction debris is crucial for their sorting and further valuable utilization. Current methods, whether relying on force sorters or separators, are reaching their limitations. In many cases, construction and demolition waste materials end up being downcycled or even sent to landfills. Our contribution is divided into two parts. The first part presents facts relating to the production of construction and demolition waste in the EU and justifies the urgency of responsible waste management. It also recaps the development of indirect methods of classification and briefly describes types of suitable sensors for material observation and algorithms for data evaluation. The second part describes our classification solution, developed by the faculties of civil engineering and electrical engineering at CTU in Prague. This method is based on the evaluation of data obtained with ordinary RGB cameras.

The dependable sorting of distinct materials from construction debris stands as a foundational element within the circular economy in construction industry. This paper introduces a technology that employs RGB cameras, mass, and acoustic sensors for material classification. By amalgamating these sensors and applying attribute extraction techniques such as shape indices, texture entropy, and intensity gradients, a comprehensive set of parameters is generated. These parameters facilitate accurate classification through the utilization of artificial intelligence.