多目标跟踪系列讲座

  • 报告人: Mahendra Mallick Principal Scientist
  • 报告人单位: Georgia Tech Research Institute (GTRI)
  • 报告地址: 研究生院研3-121
  • 报告时间: 2017年7月17、19、20、21、23日 9:00-11:30 •
  • 主题摘要:
日期 主题 7月17日 Technical Writing 7月19日 Real-world nonlinear filtering I (bearing-only and GMTI) 7月20日 Real-world nonlinear filtering II (angle-only filtering in 3D) 7月21日 Multitarget tracking and MHT I 7月23日 MHT II Lecture 1: Technical/Mathematical Writing Technical/mathematical writing is an important topic for scientists, engineers, and mathematicians for writing papers, reports, and books.Many students don’t take a course on technical/mathematical writing before completing their thesis work.This topic is vast and requires a great deal of attention. We shall present an overview of good technical writing based on recommendations by the well-known researchers/authors on the topic. Lecture 2:Three Real-world Nonlinear Filtering Problems Bearing-only filtering(BOF) in two dimensions using a single sensor or ownship is a challenging nonlinear filtering problem that has been widely studied. The BOF problem appears in submarine tracking using passive sonar measurements by an ownship, ground target tracking using acoustic sensors, and aircraft surveillance using acoustic sensors or passive radar signals. Detection, geolocation, tracking, and classification of ground moving targets in all-weather, day-night, and cluttered conditions can be performed using a ground moving target indicator (GMTI) radar. Therefore, the GMTI filtering and tracking are widely used in surveillance and precision tracking of ground movingtargets. Lecture 3:The angle-only filtering(AOF) problem in 3D using bearing and elevation angles from a single maneuvering sensor is the counterpart of the bearing-only filtering problem in 2D. The AOF problem arises in passive ranging using an infrared search and track (IRST) sensor, passive sonar, passive radar in the presence of jamming, and satellite-to-satellite passive tracking. A great deal of research has been carried out for the BOF problem in 2D. However, the number of publications for the AOF problem in 3D is relatively small.We shall present a number of filtering algorithms for BOF, GMTI filtering, and AOF problems and results of comparative evaluation of various algorithms. Lectures 4, 5:Multitarget Tracking and Multiple Hypothesis Tracking Multitarget tracking (MTT) refers to the problem of jointly estimating the number of targets and their states or trajectories. At present, the joint probabilistic data association filter (JPDAF), multiple hypothesis tracking (MHT), and the random finite set (RFS) based algorithm the generalized labeled multi-Bernoulli tracker (GLMBT) are commonly used MTT algorithms. Comparative evaluation of JPDAF and MHT show that for low probability of detection, high false alarm (FA), and closely spaced targets the MHT has significant advantage over the JPDAF. There are two types of MHT, the hypothesis-oriented MHT (HOMHT) and track-oriented MHT(TOMHT). Due to computational efficiency and software implementation, the TOMHT is preferred over HOMHT. In these lectures we shall discuss the TOMHT. A number of terms such as target, track, track hypothesis, hypothesis, global hypothesis, association hypothesis, etc. are commonly used in the MTT literature, but these terms are not clearly explained. Often a target and a track are used interchangeably. In order to remove such ambiguities we first explain these terms. Next we shall discuss the algorithm for tree-based TOMHT that uses multi-frame assignment (MFA).