Experimentation of 3D Pavement Imaging through Stereovision

Pavement distress survey plays an important role in pavement maintenance and management. Pavement distresses include items such as cracks identifiable with 2D images as well as items such as potholes identifiable with 3D images. Most of the current technologies applied in pavement survey system include 2D imaging and 1D profiler. Information on pavement 3D characteristics is not used in routine pavement surveys, primarily due to technical difficulties in hardware acquisition system and software algorithms. As the result, most of distress surveys for pavements are not automated. A novel idea proposed in the paper is to apply the stereovision technique into pavement imaging. The purpose is to reconstruct the complete 3D pavement surface from a pair of images. In this project, a total of four cameras were used in two pairs to collect pavement surface images across a 4-meter wide pavement, each pair of images covering 2 meters of the road. The overlapping of the image pixels were used to build the relations of 3D reconstruction. Four steps were involved: calibration, distortion adjustment, matching and 3-D reconstruction. Preliminary result demonstrates the feasibility of applying stereovision into pavement imaging. The accuracy of the system obtained is about 5mm at vertical direction.