Evaluation of Pavement Marking Damage Degree Based on Rotating Target Detection in Real Scenarios

Damaged road markings are widespread, and timely detection and repair of severely damaged areas is critical to the maintenance of transport infrastructure. This study proposes a method for detecting the degree of marking damage based on the top view perspective. The method improves the minimum outer rectangle detection algorithm through pavement data enhancement and multi-scale feature fusion detection head, and establishes mathematical models of different types of markings and their minimum outer rectangles to achieve accurate detection of the degree of marking damage. The experimental results show that the improved minimum bounding rectangle detection method achieves an mAP of 97.4%, which is 4.5% higher than that of the baseline model, and the minimum error in the detection of the degree of marking damage reaches 0.54%. The experimental data verified the simplicity and efficiency of the proposed method, providing important technical support for realizing large-scale road repair and maintenance in the future.