Determination of the composition of recycled aggregates using a deep learning-based image analysis

摘要: Abstract Recycled aggregates (RA) are obtained by crushing inert construction and demolition waste. Their composition is variable currently a time-consuming manual sorting. Our work makes use of deep learning, especially convolutional neural networks (CNN), to determine this in near real time an automated way. A labelled database was created for learning the CNNs. It consists approximately 36,000 images individual grains classified according their nature. After training, our best-performing CNN reaches validation accuracy 97% classifying grains. based on Residual Network that we customised order improve its performance. Moreover, evaluated mass assuming given nature have constant form density. approach compared with There less than 2% difference most RA natures tested.