A GPR Road Anomaly Interpretation System Based on YOLO Algorithm

In this paper, a ground penetrating radar (GPR) road anomaly interpretation system is designed on the basis of the YOLO object detection algorithm. The input for the system is provided by GPR field measurement image, and the GPR anomaly target interpretation is achieved through the combination of image video processing technology and YOLOv8 convolutional neural network model. The average recall rate for interpretation of GPR measured road anomaly targets attains 91.0%, while the average precision rate achieves 91.6%. Alternatively, the system is extended using the Streamlit Web App framework to design a graphical user interface (GUI), which displays images and key text information of the interpreted GPR anomalies. Ultimately, in order to verify the efficacy and feasibility of the system, explanatory experiments based on road anomaly images are conducted.