Trajectory estimation based on globally consistent homography

Siwook Nam, Hanjoo Kim, Jaihie Kim

Research output: Chapter in Book/Report/Conference proceedingChapter

2 Citations (Scopus)

Abstract

We propose a method for estimating trajectories of objects moving on a world plane. Motivation of this work is to estimate the field trajectories of players and the ball from uncalibrated monocular soccer image sequences. In order to find mappings between images and the plane, four feature points, no three of them are collinear, should exist in each image. However, many soccer images do not satisfy that condition. In that case, the object positions in the given image are mapped to those in the reference image of the sequence, and then mapped again to those in the soccer field. Conventional mapping between given image and the reference image is given by concatenation of homographies between consecutive image pairs. However, small correspondence error is accumulated in the concatenation of homographies over long image sequence. To overcome this problem, we compute globally consistent homographies for all the feature-sufficient images by solving a sparse linear system of equations which consists of consecutive and non-consecutive homographies of feature-sufficient image pairs. Experimental results with real and synthetic soccer data show that the proposed method is more accurate than existing method.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
EditorsNicolai Petkov, Michel A. Westenberg
PublisherSpringer Verlag
Pages214-221
Number of pages8
ISBN (Print)3540407308, 9783540407300
DOIs
Publication statusPublished - 2003

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume2756
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint Dive into the research topics of 'Trajectory estimation based on globally consistent homography'. Together they form a unique fingerprint.

Cite this