A HEURISTIC MULTIPLE TARGET TRACKER

Authors
  1. Beaupre, J.C.F.
  2. Farooq, M.
  3. Roy, J.M.J.
Corporate Authors
Royal Military Coll of Canada, Kingston ONT (CAN) Dept of Electrical and Computer Engineering;Defence Research Establishment Valcartier, Valcartier QUE (CAN)
Abstract
Because of the proximity of sensor returns during crisis situations such as crossing tracks, standard multiple target tracking algorithms can generate unreliable target state estimates. The multiple hypotheses tracker (MHT) is a very powerful algorithm which does not generate a single track for each target, but rather forms hypotheses composed of possible tracks and associated probability values. The hypotheses tree generated can become quite large, and by extension, very demanding in computer resources. Often, the probability values of different hypotheses are close and deciding on which set of compatible hypotheses to display on the sensor scope is difficult to make until additional sensor data becomes available. Notwithstanding these limitations in the algorithm, the human operator must always be provided with timely and accurate visual information. Knowledge elements have been formalized as heuristic rules operating on the hypotheses tree of the MHT. In effect, the MHT algorithm has bee modified to accept contextual and intrinsic knowledge in the form of heuristic rules.
Report Number
RMCC,K/O-PROC-AES-4-92-P-218 — @Paper presented at the Symposium/Workshop on Applications of Expert Systems in DND, Kingston, Ontario, 23-24 April 1992; CONTAINED IN 93-00343
Date of publication
01 Apr 1992
Number of Pages
20 (p218-237)
DSTKIM No
93-00341
CANDIS No
127661
Format(s):
Microfiche filmed at DSIS;Originator's fiche received by DSIS

Permanent link

Document 1 of 1

Date modified: