Ground penetrating radar can investigate hidden dangers such as cavities,leaks,and looseness in underground spaces,enjoying such advantages as convenience,flexibility,and simple operation.However,the amount of data obtained through large-scale ground penetrating radar detection is very large,and the processed image features are relatively com-plex.Automatically processing large-scale data and quickly screening out existing abnormal features is of great significance for real-time and rapid hidden danger investigation.This study employs the K-means clustering algorithm in unsupervised learning to filter images of common road anomalies and diseases,thereby facilitating rapid classification of anomalous data.In this experiment,the K-means algorithm demonstrated a recognition accuracy of 97.59％for samples generated by forward simulation,and reach 94.51％in the verification of actual data.Experimental results show that this method can be directly classified without labeling data features,providing a certain theoretical basis for supervised learning,and pro-viding reference value for rapid identification of underground space hidden dangers,urban road quality detection,etc.
维普
万方数据