ABSTRACT

This paper presents the design and development of a stationary microtexture road profiling system using the photometric stereo (PS) technique. The structure of the developed system is simple, mainly consisting of a digital single-lens reflex (DSLR) camera with a macro lens and multiple light-emitting diodes (LEDs). The camera with the lens is oriented perpendicularly to the pavement texture and takes images each with a different LED turned on at a time. With the pavement texture images with diverse shadings, the PS technique is applied by inverting the image-forming process locally (pixel-wise) to associate the measured image intensities with the known lighting directions to estimate the gradients for each pixel-corresponding surface patch of the pavement texture. Surface normal integration (SNI) is then employed to reconstruct the three-dimensional (3D) road surface in the microtexture scale. The PS-based system has several intrinsic advantages. First, it could achieve high accuracy for surfaces with most diffuse reflection. Second, the measurement speed is fast because of its area-scanning nature. Third, the spatial resolution is high because of the usage of a high-resolution complementary metal-oxide semiconductor DSLR camera. In addition, it can be less sensitive to effects from specularities and shadows compared with most optical-based methods, since images captured under diverse lighting directions in PS provide more cues for detection purposes. Last but not least, the hardware of the system can be made compact at low cost because of its simple structure and can be adapted for direct measurement on the pavement. Parametric studies for the Lambertian-based PS technique were first investigated analytically and numerically, and these investigations yielded the design of the system having eight LEDs with the same zenith angle of 30 degrees and uniformly distributed azimuth angles in 360 degrees. Several experimental results on various types of surfaces have demonstrated that the developed system could achieve the accuracy in the order of 10 microns.

You do not currently have access to this content.