Epipolar geometric association distance: A new cost function for passive sensor data association
Feng Ma, Huanzhang Lu, Luping Zhang, Xinglin Shen · IET Radar Sonar & Navigation · 2020
Abstract The traditional way of passive sensor data association relies on the estimation of candidate target position, and hence suffers from severe performance degradation due to the estimation error and the so‐called ghosts. The proposed new cost function is based on epipolar geometry theory that uses fundamental matrix to measure the correlation between sensor measurements directly without estimation of candidate target position. This approach has two superiorities: (1) it declines the accuracy loss caused by a large number of trigonometric operations, and (2) it greatly immunes to the impact of ghosts. A comparison with three traditional cost functions via numerical experiments demonstrates that the proposed cost function significantly improves the accuracy of data association, especially when the number of targets is more and the sensor distribution is closer.