Documents disponibles écrits par cet auteur

ANOMALY DETECTION IN CROWDED SCENES PANIC DETECTION / Hasnaa ASKOUR

Public ISBD Titre : ANOMALY DETECTION IN CROWDED SCENES PANIC DETECTION Type de document : projet fin études Auteurs : Hasnaa ASKOUR, Auteur Langues : Français (fre) Catégories : BIG DATA Index. décimale : mast 258/19 Résumé : Automate visual surveillance in public environments is a eld of research in full progress, a large number of video surveillance videos (CCTV) are used for the surveillance of busy public spaces (supermarkets, train stations, roads, streets or places pedestrians, etc.). Faced to this huge number of CCTVs, security agents can not monitor all of them at the same time. They are forced to punctually look at each of them by swapping regularly. For these reasons, this work must be automatic, means that the detection of abnormal ie "anomaly" events must be automatic for security purposes.And that last is a challenging task in Computer Vision In this project, we propose an online method for detecting Panic abnormalities in crowded video sequences. In terms of crowd behavior, Panic is the situation where there are people walking in slow pace, Then an unexpected situation happens pushing them to start running. In this project, we aim to nd these 'Panic' situations in dierent crowd scenes in the (UMN) dataset. For this purpose, we rst detect the optical ow between frames by using Farneback method. The magnitude of the dierence of two consecutive frames will give us the motion map. This last will shows the amount of motion estimated between consecutive frames. Then, by using the dierences between these motion maps, we calculate the randomness level using Entropy. By using specic thresholds to Entropy, we can detect the anomalous situations by looking if the Entropy levels exceeds the thresholds. To demonstrate and evaluate the eectiveness of the approach, we used the Receiver Operating Characteristics (ROC) curves. ANOMALY DETECTION IN CROWDED SCENES PANIC DETECTION [projet fin études] / Hasnaa ASKOUR, Auteur . - [s.d.]. Langues : Français (fre) Catégories : BIG DATA Index. décimale : mast 258/19 Résumé : Automate visual surveillance in public environments is a eld of research in full progress, a large number of video surveillance videos (CCTV) are used for the surveillance of busy public spaces (supermarkets, train stations, roads, streets or places pedestrians, etc.). Faced to this huge number of CCTVs, security agents can not monitor all of them at the same time. They are forced to punctually look at each of them by swapping regularly. For these reasons, this work must be automatic, means that the detection of abnormal ie "anomaly" events must be automatic for security purposes.And that last is a challenging task in Computer Vision In this project, we propose an online method for detecting Panic abnormalities in crowded video sequences. In terms of crowd behavior, Panic is the situation where there are people walking in slow pace, Then an unexpected situation happens pushing them to start running. In this project, we aim to nd these 'Panic' situations in dierent crowd scenes in the (UMN) dataset. For this purpose, we rst detect the optical ow between frames by using Farneback method. The magnitude of the dierence of two consecutive frames will give us the motion map. This last will shows the amount of motion estimated between consecutive frames. Then, by using the dierences between these motion maps, we calculate the randomness level using Entropy. By using specic thresholds to Entropy, we can detect the anomalous situations by looking if the Entropy levels exceeds the thresholds. To demonstrate and evaluate the eectiveness of the approach, we used the Receiver Operating Characteristics (ROC) curves.

Réservation

Réserver ce document

Exemplaires

Code barre	Cote	Support	Localisation	Section	Disponibilité
mast 258/19	mast 258/19 HAS	Texte imprimé	Unité des masters	Mast/19	Disponible