Hawkmail
Simulators & Vehicles
Research
Media & Events
Staff
| Document Number: | N2015-001 |
| Document Type: | Project report |
| Author(s): | Vincent Horosewski Romeo Abraham Shawn Allen Dawn Marshall |
| Publication / Venue Name: | None listed |
| Publication Date: | 2015-02-18 |
| Abstract: | This report includes information on methods or algorithms used in automating the extraction of road vector data, which resulted in the improvement of road vector data extraction from aerial imagery. In the current phase of work, two methods were explored, a method based on existing GIS tools and one based on image processing techniques. Existing GIS tools can successfully extract road vector data from images of arbitrary complexity. However, this analysis requires significant manual refinement of intermediate results in order to obtain good results. Image processing techniques can be leveraged to automate much of the process of road vector extraction. Obtaining good results, however, can be challenging on more complicated imagery. Additionally, complex analyses can require minutes per image, although this does not require human oversight. Exploration of this method indicates that some means of differentiating between roads and similar-looking features must be incorporated to fully automate the process of road vector extraction. While presented separately, these two methods can be viewed as complementary approaches. That is, image processing techniques can be applied to imagery initially, and the results obtained can be plugged into GIS tools for refinement and additional processing. In this way, time spent on manually cleaning up the images will be reduced. |
| Body: |  |
| Copyright: | NADS |
| Keywords: | road network visualization raster vector data |
