Automated Extraction of Horizontal Curve Information for Low-Volume Roads. Issue 1 (January 2015)

• Record Type: Journal Article
• Authors: Li, Zhixia; Chitturi, Madhav V.; Bill, Andrea R.; Zheng, Dongxi; Noyce, David A.
• Is Part Of: Transportation research record. Volume 2472:Issue 1(2015)
• Page Start: 172
• Page End: 184
• Publication Date: 2015-01
• Subjects: Transportation -- Periodicals; Roads; Transport -- Périodiques; Routes -- Périodiques; Routes -- Conception et construction -- Périodiques; Roads; Transportation; 388.05
• DOI: 10.3141/2472-20
• Languages: English
• ISSNs: 0361-1981

Abstract

Rates of fatal and injury crashes on low-volume roads are much higher than those on higher-volume roads. A large proportion of crashes on low-volume roads are roadway departure crashes. Research has shown that roadway departure crashes are 1.5 to 4 times more likely at curves than on tangent sections. Therefore, knowledge of where horizontal curves are located and their geometric characteristics is key to addressing safety issues for low-volume roads. In the United States, although many states have a horizontal curve database for their state routes and Interstate highways, most states do not have such a database for non-state-owned roads, especially for low-volume rural roads, because of the lack of funds for additional data collection. A novel approach is presented that can extract curve data for both state and local roads in an accurate, cost-effective, and time-efficient manner without additional data collection. An ArcGIS add-in tool, CurveFinder, was developed by the authors previously to automatically identify horizontal curves from a selected roadway layer, classify curves, compute curve geometrics, and, finally, create a geographic information system for curve layers. CurveFinder was updated to incorporate compatibility with curve elements of FHWA's Model Inventory of Roadway Elements. Case studies are used to demonstrate the application of the updated CurveFinder on rural low-volume roads in different states and the lessons learned. Results show that none of the control curves were 100% missed by CurveFinder. False identification rates were also low. Both results validate CurveFinder. Causes of the few errors in the extracted curve data are identified, and potential solutions are explored.