Road network extraction and uncertainty analysis

Road maps can provide information on not only topographical conditions but also on how materials and people move. Historically, available information pertaining to road maps consisted of static images of preordained routes. Even now, Google maps and hand held Global Positioning Systems (GPS) represent a static view of road networks, requiring either recapturing images or manually updating units. However, in order to have more current, information rich representations of transportation networks or road maps, the use of kinematic information provided by sensors such as Ground Moving Target Indicator (GMTI) data is required. The data these GMTI sensors supply is not only a static representation of a target but also the kinematics. The approach employed for synthesizing the received data into a complete estimate of the road network is through the use of the Hough Transform, to identify line segments which collectively represent the road network. The Total Least Squares is used to characterize the uncertainty associated with this representation as well as provide a more accurate estimate. The uncertainty in each of the estimated parameters (i.e. slope and intercept of a line) can then be approximated by the Cramer Rao lower bounds. Finally the identified segments are merged and connected to provide a more complete representation of the road network. The approach used here is iterative, for example, when new data within the area of interest is received, a better estimate of the road segment can be obtained and the overall road network is updated and allowed to grow in all directions.