Measurement of Pavement 3D Texture Using the Binocular Reconstruction Algorithm Improved by Moving Laser Line Constraint

Abstract

For such a specific object as asphalt pavement with inconspicuous feature points, the traditional image vision algorithm has limited effect on its texture measurement. In order to realize the high-resolution, high-precision, and convenient detection of three-dimensional (3D) pavement texture while better serving traffic safety, the traditional binocular reconstruction technology was improved in this study threefold. First, the improved binocular reconstruction test system and the measurement accuracy evaluation device were manufactured to realize the reconstruction of the 3D texture and the true point-to-point evaluation of measurement accuracy. Second, the global scanning constraint posed by a moving laser line was introduced to shoot video images, which formed numerous mandatory constraints and improved the matching accuracy. Last, the centroid enhanced subregion segmentation algorithm was proposed to complete stereo matching under global scanning constraint. Results show that the binocular reconstruction algorithm improved by moving a laser line constraint under a centroid enhanced subregion segmentation matching mode can significantly improve the accuracy in the measurement of asphalt pavement 3D texture. Ultimately, the test results can meet the precision requirement of micro-texture.