A Novel Method for Extracting Road Map from The Historical Measurement Set of Sensors

Abstract

The road information hidden in the measurements of sensors can be mined to extract the road map and improve the tracking accuracy of the subsequent ground target. In this paper, a novel method which makes full use of the historical measurement set (HMS) of sensors to extract road map is proposed. Firstly, in the case of dense target situations, missing detections and false alarms, the Gaussian mixture probability hypothesis density (GMPHD) filter is a method at choice to estimate the multi-target state, and the multi-target state for a period of time constitutes the multi-target historical state set (MTHSS). Secondly, a density-based clustering method, DBSCAN, is performed on the MTHSS to obtain the road feature points. Finally, the key road nodes are extracted through road feature points by the Douglas-Peukcer (D-P) method. The simulation results show that the proposed method can extract the road map from the HMS of sensors effectively, and the accuracy of the extracted road is sufficient to assist in ground target tracking.