Accurate and Less Expensive Pavement Distress Surveying Using Multi-Photogrammetric-Output Fusion

Automatic pavement distress detection, data collection and mapping typically relies on line-scan cameras and image processing technologies, while rutting is mainly extracted using laser profilers scanning along the wheel paths. Two dimensional image processing based systems still lack robust sensor modeling; hence robust detection and modeling of accurate and detailed pavement distress data has not yet been met after decades of 2D image analysis research. On the other hand, laser scanning is very expensive and less portable than typical imaging systems. Thus the arrays of laser profilers for 3D data acquisition have limitations for enhancing or replacing image-based outputs. As a robust approach to overcome some of the above current systems limitations, possibly through integration, close-range digital photogrammetry is tested with multiple types of pavement distresses including cracks, rutting, and potholes. Tested samples showed an accurate reconstruction of 3D pavement surfaces including all types of pavement distresses. The outputs include: 1) 3D point-clouds accurately generated retaining the natural surface color and texture information for friendly human, vision data analysis, or for machine vision analysis, 2) Triangular Irregular Networks (TIN) of the pavement surfaces, 3) Profiles, and 4) 2D and 3D contour lines. These outputs add richness to the information, and by eliminating the deceptive aspects of artifacts such as filled cracks, the 3D output should improve the performance of existing automated vision algorithms. Fusion of different outputs also facilitated experiments in detection and mapping of cracks, texture (for non-contact friction and noise measurement), potholes, rutting and other types of pavement distresses which are briefly presented here. It is noteworthy to mention that all the experimental work was carried out using a consumer-grade DSLR camera, commercial software, and only a few custom programs. The built-in flash and natural daylight were the only sources of illumination.