Optical masonry waste analysis using machine learning

Abstract

This paper presents the results of recent research analysis of masonry waste, combining optical sensor technology with machine learning methods. A material dataset was created from various masonry and brick fractions and categorized based on water absorption and bulk density. Using these materials, an image dataset was generated with a high‐resolution RGB line‐scan camera. The acquired images were processed and characteristic features were extracted. This data formed the basis for the selection and training of a classification model, which was subsequently implemented on a demonstrator system. This system records and analyzes bulk material in free fall, enabling a process‐oriented investigation. The trained model was validated using newly acquired image data of known categories under continuous operating conditions. The achieved recognition rates demonstrate that classification of different brick and masonry types based on image features is feasible.