Road Network Estimation Using Implicit Curves

Abstract

Airborne Ground Moving Target Indicator (GMTI) measurements are used for generating and analyzing traffic flow. GMTI generated tracks can be of poor quality because of false returns, doppler blindness etc. Knowledge of road networks is often used to enhance the quality of the kinematic estimates of the position of vehicles on the road. In this work, the focus is on developing or improving the estimate of a road network based on track estimates generated by GMTI tracks. Assuming that the road network can be discretized into multiple segments which are characterized by straight lines, ellipses or arcs of circles, a sparse parameterization of a road network can be synthesized. Hough transform of all available track data is used to sequentially identify straight line segments followed by ellipses. After sorting the track data into bins for straight line segments and ellipses, an eigenvalue/eigenvector based approach is used to identify the mean and covariance of the parameters of the polynomial curves. The Kanatani-Cramer-Rao lower bounds are also derived to quantify the quality of the estimate. Hough transform based sorted data is used to illustrate the proposed estimation technique for a straight line segment of a road and a roundabout like road segment.