Search

Horizontal roadway curvature data are essential for roadway safety analysis. However, collecting such data is time-consuming, costly, and dangerous using traditional, manual surveying methods. It is especially difficult to perform such manual measurement when roadways have high traffic volumes. Thus, it would be valuable for transportation agencies if roadway curvature data could be computed from photographic images taken using low-cost digital cameras. This is the first article that develops an algorithm using emerging vision technology to acquire horizontal roadway curvature data from roadway images to perform roadway safety assessment. The proposed algorithm consists of four steps: (1) curve edges image processing, (2) mapping edge positions from an image domain to the real-world domain, (3) calibrating camera parameters, and (4) calculating the curve radius and center from curve points. The proposed algorithm was tested on roadways having various levels of curves and using different image sources to demonstrate its capability. The ground truth curvatures for two cases were also collected to evaluate the error of the proposed algorithm. The test results are very promising, and the computed curvatures are especially accurate for curves of small radii (less than 66 m/200 ft) with less than 1.0% relative errors with respect to the ground truth data. The proposed algorithm can be used as an alternative method that complements the traditional measurement methods used by state DOTs to collect roadway curvature data. [ABSTRACT FROM AUTHOR]