Tracking Video Target via Particle Filtering on Manifold
| Autor: | Huilin Ge, Kang Lou, Zhiyu Zhu |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Geodesic
Euclidean space Covariance matrix Computer science business.industry ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Lie group 020206 networking & telecommunications 02 engineering and technology Covariance Computer Science Applications Control and Systems Engineering Robustness (computer science) 020204 information systems 0202 electrical engineering electronic engineering information engineering Computer vision Affine transformation Artificial intelligence Electrical and Electronic Engineering Particle filter business |
| Zdroj: | Information Technology and Control. 48:538-544 |
| ISSN: | 2335-884X 1392-124X |
| DOI: | 10.5755/j01.itc.48.4.23939 |
| Popis: | Most of existing particle filtering-based video target tracking algorithms are in Euclidean space, when object posture and scale size changes, and to track high dimensional system, it is difficult to guarantee the tracking effect. This paper describes the covariance descriptor to represent the object image region, the geometric deformation of the object image region can be realized by an affine transformation, and the affine transformation matrix is one element of the Lie group. Then particle filter algorithm based on lie group of manifold is proposed , the video tracking system state lies directly on a low dimensional manifold, state samples are drawn moving on the manifold geodesics, thus state space of intrinsic geometrical characteristic can be in full use, which provides a new idea for improving the tracking efficiency and robustness. Simulation results show that object in the case of geometric deformation including scale size changes, rotating, etc. The proposed manifold particle filtering algorithm can still realize target tracking well and improve the real-time performance. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|