Tracklet and Signature Representation Using Part Appearance Mixture Approach in the Context of Multi-shot Person Re-Identification
| Autor: | Awatef Ben Fradj, Abdennaceur Kachouri, Mohamed Amine Ben Farah, Francois Bremond, Salwa Baabou, Furqan M. Khan |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Computer science
Orientation (computer vision) business.industry ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Representation (systemics) Context (language use) Pattern recognition Space (commercial competition) Mixture model Identifier Feature (computer vision) Artificial intelligence Set (psychology) business |
| Zdroj: | Advanced Methods for Human Biometrics ISBN: 9783030819811 |
| DOI: | 10.1007/978-3-030-81982-8_7 |
| Popis: | Recognizing persons in a video surveillance scene in the real world is attractive and is now showing an increasing interest. The task of person Re-Identification (Re-ID) consists in assigning the same identifier to all instances of a particular individual captured in a series of images or videos, even after the occurrence of significant gaps over time or space. This chapter addresses the problem of high variance in a person’s appearance by represent it as a set of multi-modal feature distributions modeled by Gaussian Mixture Model (GMM). We tackle also the Re-ID challenges that consist in occlusions, illumination changes and person’s orientation/pose. To this end, we extract the tracklets, i.e trajectories of persons. From these tracklets, we compute the signature and represent it based on the approach of Part Appearance Mixture (PAM) in the context of Multi-Shot person Re-Identification. Experiments and results on two public datasets and on our own dataset show good performance. An evaluation of the quality of this signature representation is also described. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|