Particle Size Detection Of Sand And Gravel Images Based On Deep Learning

With the rapid development of the civil engineering industry,the demand for concrete in various construction industries at home and abroad has continued to increase,and the demand for sandstone as one of the main components of concrete is also increasing.The concrete used in different applications has different requirements for the size of the sandstone particles.How to intelligently detect the particle size of sandstone and improve the production efficiency of the civil engineering industry is an urgent research problem.From the perspective of image processing,on the basis of analyzing the problems of particle size detection using traditional methods,according to the characteristics of clone adhesion and mutual occlusion of sandstone particles,this dissertation puts forward a method of sandstone image particle size detection based on two-stage deep learning.In the first stage,a sandstone segmentation network model was constructed to initially segment the sandstone target from the invalid background.The model is built on a fully convolutional network,uses the multiple convolutional layers to extract the features of sandstone image.After the corrosion expansion and morphological operation,the initial segmentation result is obtained.This is the segmentation phase.In the second stage,a separate network model similar to the segmentation model was used to separate the tightly adhered and blocked sandstone targets from each other to obtain each independent sandstone target,which is the separation stage.After the sandstone separation,the longest diameter of the sandstone target is used as the particle size of the sandstone.The average particle size of the sandstone image is obtained by removing the invalid value and then averaging.Through experiments,our method can obtain the average particle size of the sandstone target quickly and accurately.In terms of sandstone image segmentation ability,compared with other existing segmentation algorithms,our method can accurately and effectively separate the closely adhered sandstone targets,and improve the accuracy of sandstone target particle size calculation.