Association using modified Global Nearest Neighbor in the presence of bias
Mengzhao Shi, Qiang Ling, Zhaohua Yu, Jin Zhu · Chinese Control Conference · 2013
Global Nearest Neighbor (GNN) method has been widely implemented in multi-target multi-sensor tracking system, and has achieved a good tracking performance. However, the performance of this approach can be easily destroyed when the sensor biases are involved in the target observations, especially in the cluttered environment with false alarms and missed detections. In order to tackle this issue, this paper proposes a modified GNN method to associate observations and targets in the presence of the sensor bias. Compared with the traditional GNN method, the modified GNN method introduces the new concept of target pattern, and computes the similarity between observations using the feature vector which is constructed using the distance among observations. Simulations show that the modified GNN method has high association success rate, and is robust against the variation of the sensor bias, the number of targets, and the clutter density which the GNN method cannot handle well.