Combining Image Processing with Signal Processing to Improve Radio Position Estimation

This paper uses aerial, orthorectified imagery to model the radio multipath environment. This model is used to improve time difference of arrival (TDOA) based radio positioning. The imagery is processed using a building extraction algorithm compiled from current techniques. The building model is used to develop a shortest-path routing protocol from each candidate transmitter position to each receiver, creating more accurate models of the TDOA information than are available when assuming all paths are line of sight (LOS). The final positioning algorithm is a maximum likelihood algorithm that compares the observed and theoretical TDOA values at each grid point. When compared to the Chan & Ho method which assumes line of sight (LOS), the method in this paper improves transmitter geolocation error by an average of 44 m (53%) in non-LOS environments. However, in cases where all receivers actually have a LOS, the current method is faster and slightly more accurate. Therefore, the method in this paper is most applicable to scenarios requiring position estimation of vehicles in an urban environment using stationary receivers and significant computing power.