Evaluating Remotely Sensed Images for use in Inventorying Roadway Infrastructure Features

In order to effectively manage and maintain the nation’s transportation infrastructure, accurate inventory and condition data are required. Up-to-date information helps to identify safety or operationally deficient elements, prioritize maintenance needs, and monitor conditions (Opiela and Perkins, 1986). As a result, all Departments of Transportation (DOTs) in the United States maintain some type of roadway feature inventory (Karimi et al., 2000). Roadway inventory data are used by DOTs for a variety of purposes including traffic safety, construction projects, traffic engineering studies, evaluation of maintenance needs, and planning. Inventory data are also collected and maintained to meet Federal data reporting requirements. Infrastructure data are used by the Federal Highway Administration (FHWA) as informational support for the “Condition and Performance Report” to Congress, as well as data that appears in various FHWA publications. Inventory data are also used by other agencies including other state and local agencies, business and industry, educational institutions, the media, and the general public (FHWA, 2000). Inventory data are necessary to support the numerous functions within DOTs and state and local transportation agencies that utilize those data. However, most data collection methods, which include manual methods, global positioning systems (GPS), and video or photolog vans, are conducted in the field requiring significant time and resources to cover even a minor amount of roadway. This is problematic since both state DOTs and local areas are responsible for significant street network systems. Iowa, for example, has a surface street system covering approximately 110,000 linear miles. As a result, resource constraints often dictate that only minimal inventory data elements, such as pavement condition or number of lanes, are collected and reported systemwide. Other data are collected at the corridor level as needed for specific uses such as planning of new construction or evaluation of high accident locations. To meet data and reporting needs, sampling of subsets of roadway segments are often used and then extrapolated to provide systemwide estimates. While sampling may provide adequate information for some uses, other types of applications or analyses are limited by the inability to cost-effectively collect comprehensive data. Traffic studies, safety studies, and evaluation of access control require more comprehensive data than is provided by sampling. Safety studies in particular could benefit from more data. Many aspects of the roadway have been correlated to occurrence of accidents. For example, a narrow bridge width to approaching roadway width ratio has been associated with increase in accident (Fitzpatrick et al., 2000). Other roadway features such as lane width, shoulder width, and location of utility poles or other fixed objects along the roadway influence likelihood and severity of accidents. However, even with crash data that has been accurately spatially located, it is almost impossible to evaluate roadway deficiencies without a supporting database of roadway information