A hybrid system for affine-invariant trajectory retrieval
Authors:
Open Access
- 15 October 2004
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 235-242
Abstract
This paper studies efficient feature spaces for content based indexing and retrieval of object motion trajectories. Taking object trajectory data as input, we first investigate highly compact affine invariant feature spaces based on Fourier Descriptors (FD) and Principal Component Analysis (PCA) techniques. Based on these feature spaces, we then develop a hybrid content based indexing and retrieval system that employs a two-stage matching scheme. The first stage uses affine-invariant Fourier Descriptor (FD) for indexing and retrieval. Top few results from this stage along with the original query are then posed to the second stage of the matching system that employs Principal Component Analysis (PCA) for fast retrieval. We compare our system's performance with two other approaches borrowed from 2-D shape representation in image analysis. For quantitative analysis of the system performance, we report query results in terms of precision-recall metricsKeywords
This publication has 9 references indexed in Scilit:
- Segmented trajectory based indexing and retrieval of video data2004
- A comparative study of curvature scale space and Fourier descriptors for shape-based image retrievalJournal of Visual Communication and Image Representation, 2003
- Modal matching for correspondence and recognitionIEEE Transactions on Pattern Analysis and Machine Intelligence, 1995
- Matching motion trajectories using scale-spacePattern Recognition, 1993
- A theory of multiscale, curvature-based shape representation for planar curvesIEEE Transactions on Pattern Analysis and Machine Intelligence, 1992
- Application of affine-invariant Fourier descriptors to recognition of 3-D objectsIEEE Transactions on Pattern Analysis and Machine Intelligence, 1990
- A critical investigation of recall and precision as measures of retrieval system performanceACM Transactions on Information Systems, 1989
- Principal Component Analysis1986
- Asymptotic eigenfunctions and eigenvalues of a homogeneous integral equationIEEE Transactions on Information Theory, 1962