Skip to Main content Skip to Navigation
Theses

Instantiation of a textual description schema of video surveillance scenes

Abstract : Surveillance systems are important tools for law enforcement agencies for fighting crimes. Surveillance control rooms have two main duties: live monitoring the surveillance areas, and crime solving by investigating the archives. To support these difficult tasks, several significant solutions from the research and market fields have been proposed. However, the lack of generic and precise models for video content representation make the building of fully automated intelligent video analysis and description system a challenging task. Furthermore, the application domain still shows a big gap between the research field and the real practical needs, it also shows a lack between these real needs and the on-market video analytics tools. Consequently, in conventional surveillance systems, live monitoring and investigating the archives still rely mostly on human operators. This thesis proposes a novel approach for textual describing important contents in videos surveillance scenes, based on new generic context-free "VSSD ontology", with focus on two objects interactions. The proposed ontology presents a new generic flexible and extensible ontology dedicated for video surveillance scenes description. While analysing and understanding variety of video scenes, our approach introduces many new concepts and methods concerning mediation and action at a distant, abstraction in the description, and a new manner of categorizing the scenes. It introduces a new heuristic way to discriminate between deformable and non-deformable objects in the scenes. It also highlights and exports important features for better video objects interactions learning classifications and for better description. These features, if used as key parameters in video analytics tools, are much suitable for supporting surveillance systems operators through generating alerts, and intelligent search. Moreover, our system outputs can support police incidents reports, according to investigators needs, with many types of automatic textual description based on new well-structured rule-based schemas or templates. [...]
Complete list of metadatas

Cited literature [365 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-02965857
Contributor : Abes Star :  Contact
Submitted on : Tuesday, October 13, 2020 - 3:36:08 PM
Last modification on : Thursday, October 15, 2020 - 4:05:32 AM

File

2019TOU30249a.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02965857, version 1

Citation

Wael Farid Youssef. Instantiation of a textual description schema of video surveillance scenes. Image Processing [eess.IV]. Université Paul Sabatier - Toulouse III, 2019. English. ⟨NNT : 2019TOU30249⟩. ⟨tel-02965857⟩

Share

Metrics

Record views

43

Files downloads

15