Abstract: A method for identifying and extracting characterization parameters of recycled concrete sand particles based on deep learning technology is provided. The method integrates image processing method based on deep learning and quickly recognition of recycled concrete sand particles (RCSP), adopts U-Net semantic segmentation model, develops RCSP data set by inventing a 3D image acquisition platform equipment of recycled concrete sand, in which two CCD industrial cameras are used to collect original multi-dimensional images of the moving RCSP synchronously in the same frame. Secondly, data sets are separated into training set and verification set by 4:1, in which training set are first used to train the U-Net semantic segmentation model to quickly identify the recycled concrete sand, during this process the best training parameters of U-Net semantic segmentation model are determined. Finally, the verification sets are adopted to validate the training model.

Abstract: A method for identifying and extracting characterization parameters of recycled concrete sand particles based on deep learning technology is provided. The method integrates image processing method based on deep learning and quickly recognition of recycled concrete sand particles (RCSP), adopts U-Net semantic segmentation model, develops RCSP data set by inventing a 3D image acquisition platform equipment of recycled concrete sand, in which two CCD industrial cameras are used to collect original multi-dimensional images of the moving RCSP synchronously in the same frame. Secondly, data sets are separated into training set and verification set by 4:1, in which training set are first used to train the U-Net semantic segmentation model to quickly identify the recycled concrete sand, during this process the best training parameters of U-Net semantic segmentation model are determined. Finally, the verification sets are adopted to validate the training model.