Multi-region perspective transformation-based measurement of pavement disease size

To improve the efficiency and accuracy of pavement disease inspection and measurement, this study investigates the physical dimension measurement of pavement disease based on pixel coordinates identified by machine vision. To address the increased measurement error caused by nonlinear image distortion, a multi-region perspective transformation method is proposed for disease size measurement. Calibration images were collected through simulation calibration experiments, and 2 methods for computing perspective transformation matrices were compared using reprojection error. The OpenCV method was selected for multi-region perspective transformation, and the pixel-to-physical dimension conversion ratio coefficients were calculated to establish a measurement parameter library. Further refinement of coordinate transformation and size calculation by region significantly improved measurement accuracy. Experimental results show that measurement accuracy exceeds 96% in the near field (0–6 m), reaches over 90% in the mid-field (6– 10 m), and remains above 85% in the far field (10–18 m). This method significantly enhances the reliability of pavement disease size measurements and provides effective technical support for intelligent pavement maintenance.