Research On Sand And Gravel Aggregate Quality Detection Based On Image Processing And Deep Learning

The quality of concrete is the guarantee of engineering buildings,and it’s depends on the quality of concrete raw materials.The quality detection of sand and gravel aggregate can’t be ignored,because of it’s one of the important components of concrete and high quality is the guarantee of concrete strength and hardness.The traditional sand and stone aggregate quality detection efficiency is low,it’s unable to accurately and effectively detect the quality of sand and stone aggregate because of affected by human factors.This paper proposes a method based on image processing and deep learning to detect the quality of sand and gravel aggregate.The image processing algorithm and deep learning algorithm are integrated to detect the quality of sand and gravel aggregate accurately,and the concrete aggregate quality detection system is designed to complete the intellectualization and automation of sand and gravel aggregate quality detection.When the image acquisition device of the concrete aggregate intelligent detection system completes the image acquisition of the sand and gravel aggregate,the deep learning algorithm evaluates the quality grade of the sand and gravel aggregate,and then analyzes its characteristic parameters by the image processing algorithm.Finally,the quality grade and characteristic parameters of the sand and gravel aggregate are displayed in the man-machine interface.The main research work of this paper is as follows.First,it’s to evaluate the quality grade of sand and gravel aggregate.The quality grade of sand and gravel aggregate in concrete production enterprises is evaluated,and 5000 sand and gravel aggregate images are collected.According to the characteristics of sand and gravel aggregate images,the depth convolution neural network model CNN13 is builted.Based on the aggregate images data set,back propagation algorithm,adaptive moment estimation optimization and GPU high-speed training are used to train and test on VGG16 model and CNN13 model respectively.The accuracy of CNN13 model is 98.1%,which is3.9% higher than VGG16 is 94.2%.Secondly,through the image processing algorithm,including image filtering,image enhancement,image segmentation,morphological processing and other image processing technologies,the sand and gravel aggregate contour is extracted from the sand and gravel aggregate image.Convex hull fitting is used to approach the sand and gravel aggregate contour to the greatest extent,and then the area pixel parameters,particle size parameters and particle shape parameters of sand and gravel aggregate are extracted.Then design the sand and gravel aggregate quality detection system,according to the complex lighting conditions of sand and gravel aggregate image acquisition,through the selection of appropriate camera,lens,light source,design a sealed image acquisition device,minimize the external light impact in the process of sand and gravel aggregate image acquisition,and eliminate the distortion of sand and gravel aggregate image in the process of acquisition through camera calibration change the problem.Finally,the image processing algorithm and deep learning algorithm are integrated,and the concrete aggregate intelligent detection system is designed through the man-machine interface with the aggregate collection device,which improves the process of aggregate quality detection,and completes the intellectualization and automation of aggregate quality detection process.