A Data Fusion Approach for Extracting Highway Maintenance Features

Utah Department of Transportation's (UDOT’s) Maintenance Feature Inventory primarily relies on the mobile Light Detection and Ranging (LiDAR) data collected by Mandli Communications. In a recent UDOT project, the research team at Utah State University (USU) investigated whether aerial LiDAR data can be a more cost-effective means than the mobile counterparts to identify highway features in areas where large-scale changes may have occurred, and a way to keep the feature inventory current. In the airborne data collection conducted by USU in 2015, both aerial LiDAR point cloud and high-resolution aerial imagery data were obtained. Four highway sections in Utah were covered: one section on I-84, two sections on I-15, and one section on US-191. Preliminary results show that aerial LiDAR is a promising technology in detecting some highway features, such as guardrails, medians, and light poles, as well as large road signs. In particular, aerial LiDAR data provide a view of the roadway from a different perspective. As a result, it allows features that may have been hidden from the mobile platform to be viewed and identified. For example, culverts and bridges can be effectively detected using aerial LiDAR data. However, due to the lower data resolution, smaller highway features, such as speed limit signs, are difficult to be identified accurately by aerial LiDAR data. In the previous study, the project team also found that georeferenced high-resolution aerial imagery data can be used as an alternative means to identify highway features. Moreover, aerial imagery data are particularly suitable for developing automated procedures for highway feature extraction, which offers great potential in time and cost savings when updating the feature inventory. With the advent of Unmanned Aerial Vehicle (UAV) technology, high-resolution aerial images will be much more affordable and easily accessible for transportation agencies in the future. Although the current multispectral aerial image data set was collected with a fixed-wing plane, the methodology developed for the current data set will be readily transferable to any UAV-based data collection platform. However, aerial imagery data do not contain any elevation information, which is critical for identifying some highway features. To achieve the maximum level of accuracy and completeness, it is imperative to develop a data fusion approach that utilizes both aerial LiDAR and aerial imagery data to compensate for the limitations of both methods. Also, automated feature extraction is an important aspect of large scale applications. Therefore, the project team proposes to conduct a study to take advantage of the unique data set obtained from the recent UDOT study by developing an efficient workflow for automatic extraction of highway maintenance features. Most studies investigating highway feature extraction in the literature focus on a single source of data, e.g., aerial LiDAR or imagery data. Based on the project team's previous studies, the project team found that each data collection method has its advantages and limitations. The most effective approach to achieve the maximum level of accuracy and completeness is probably to combine data collected from multiple sources. The data set obtained from the project team's recent UDOT study contains both aerial LiDAR and aerial imagery data, and provides us with a unique opportunity to pursue this research direction. The research objectives are to: (1) Develop an efficient workflow that utilizes both aerial LiDAR and aerial imagery data to compensate for the limitations of both methods; and (2) Propose an automated feature extraction procedure for large scale applications. This will be accomplished by: (1) Conduct a comprehensive literature review of highway feature extraction using aerial LiDAR and imagery data; (2) Develop an image processing algorithm to process the high-resolution aerial imagery data; (3) Propose a data fusion procedure that combines the advantages of both data sets; and (4) Evaluate the effectiveness of the data fusion approach by comparing the accuracy and completeness of highway features with the previous study. The expected results are to develop: (1) An efficient workflow that utilizes both aerial LiDAR and aerial imagery data; and (2) An automated feature extraction procedure for large scale applications.

    Language

    • English

    Project

    • Status: Programmed
    • Funding: $100000
    • Sponsor Organizations:

      Utah Department of Transportation

      4501 South 2700 West
      Project Development
      Salt Lake City, UT  United States  84114-8380
    • Managing Organizations:

      Utah Department of Transportation

      4501 South 2700 West
      Project Development
      Salt Lake City, UT  United States  84114-8380
    • Project Managers:

      Hales, Thomas

    • Performing Organizations:

      Utah State University, Logan

      Civil and Environmental Engineering Department
      Logan, UT  United States  84332
    • Principal Investigators:

      Song, Ziqi

    • Start Date: 20161003
    • Expected Completion Date: 20171031
    • Actual Completion Date: 0

    Subject/Index Terms

    Filing Info

    • Accession Number: 01607617
    • Record Type: Research project
    • Source Agency: Utah Department of Transportation
    • Files: RiP, STATEDOT
    • Created Date: Aug 12 2016 3:07PM