Analiza greške u algoritmima za praćenje pokretnih ciljeva primenom algoritma GNN / A study of a target tracking method using Global Nearest Neighbor algorithm

Abstract

Izbor podesnog algoritma u skladu sa raspoloživim računarskim resursima važna je karika u lancu otkrivanja i praćenja pokretnih ciljeva pomoću osmatračkog radara. Efikasnost ovih algoritama zavisi od izbora estimatora stanja, položaja i brzine i metode za asocijaciju podataka. U radu je analizirana greška praćenja 20 ciljeva pomoću algoritma "najbližeg suseda " (GNN). Za estimaciju stanja korišćenje Kalmanov filter i algoritam IMM. Mera kvaliteta praćenja je srednja kvadratna greška po poziciji. Za simulaciju praćenja odabrane su putanje leta četiri grupe po pet ciljeva sa i bez manevra. / This paper compares two algorithm for Multiple Target Tracking using Global Nearest Neighbor (GNN) approach: first by the use standard Kalman filter (SKF-GNN) and second by the use Interacting Multiple Model (IMMGNN) in order to derive final tracking estimation. For both algorithms the observations are divided in clusters to reduce computational efforts. Results of simulations by tracking 20 targets simultaneously reveal that in some cases the IMMGNN approach gives better solution then KF-GNN approach. The computational time for assignment problem solution for maneuvering target tracking using both algorithm is studied and results prove that is IMMGNN suitable for real time implementations.

Keywords

• praćenje pokretnih ciljeva
• Kalmanov filter
• algoritmi IMM i GNN
• multiple target tracking
• data association
• Global Nearest Neighbor