Road Map Extraction Using GMPHD Filter and Linear Regression Method for Ground Target Tracking

Abstract

The effective use of road map information can greatly improve the ground target tracking performance, but in the cases where the road map is not available, the road map can be extracted through the target state moving on the road. Therefore, this paper has proposed a novel road map extraction algorithm in the Random Finite Set (RFS) formulation. This novel algorithm exploits the Gaussian Mixture Probability Hypothesis Density (GMPHD) filter and linear regression method to extract the road map. First of all, the single target state is estimated by the GMPHD filter in clutter, and the estimation of multiple-target state in continuous observation time constitutes the historical data of multi-target state. Then, the linear regression method is performed on these data. The linear model is the road model, and the training set is the historical data of multi-target state. After the above two steps, the road map can be extracted. The performance of the proposed algorithm is validated through the road map extraction simulation. The simulation results demonstrate that compared with the existing methods, the precision of the extracted road segment can be improved significantly through the proposed algorithm. Further, the extracted road is used for the ground target tracking. The simulation results indicate that the road information extracted by the algorithm proposed in this paper can effectively improve the tracking precision of moving targets on the road.