In view of the issues of large excavation volumes in the current Pinglu Canal sand excavation project,the long time required for laboratory testing of mud content,and the complex procedures involved,a rapid detection technique for the mud content in sand was proposed,based on quick slurry preparation and image recognition.The method involves sampling,preparing the slurry and taking a timed image of the sample after sedimentation.The color features extracted from the test area and the black-and-white comparison area are used as inputs,with mud content as the output.A database of 254 data sets is established,with mud content as the output.Using artificial neural networks,mud content prediction training and hyperparameter optimization are carried out.The comparison between the prediction results of the test set and the actual results show that the method has high accuracy for detecting low mud content(below 10％),with an error within 1％.For high mud content detection,the average error is within 5％.This technique simplifies the process,improves detection efficiency,and enables real-time transportation flow decision-making on site(direct resource utilization,use after washing,or landfill disposal),significantly enhancing construction efficiency.
维普
万方数据
