Visual surveillance of human activity

2011

Visual surveillance in dynamic scenes, especially for human and some objects is one of the most active research areas. An attempt has been made to this issue in this work. It has wide spectrum of promising application including human identification to detect the suspicious behavior, crowd flux statistics, and congestion analysis using multiple cameras. In this paper deals with the problem of detecting and tracking multiple moving people in a static background. Detection of foreground object is done by background subtraction. Detected objects are identified and analyzed through different blobs. Then tracking is performed by matching corresponding features of blob. An algorithm has been developed in this perspective using Angular Deviation of Center of Gravity (ADCG), which gives a satisfying result for segmentation of human object.

Détection, suivi, et analyse de comportement des personnes en mouvement dans les systèmes de vidéo surveillance : une approche basée sur les graphes Résumé Dans les dernières deux décennies nous avons assisté à un rapide accroissement de la disponibilité d'informations audio-visulelles dans différents domaines. Suite à cette diffusion, on a enregistré un considérable intérêt de la part de la communauté scientifique dans l'approfondissement des techniques de vision artificielle. Un des domaines dans lesquels la recherche scientifique s'est particulièrement investie, a été celui de l'analyse vidéo, en temps réel, pour la détection d'évènements d'intérêt présents dans la scène contrôlée par la caméra. La détection d'événements dans une séquence vidéo est appliquée à différents domaines applicatifs, le plus important étant les systèmes de vidéo surveillance. Les systèmes de vidéo surveillance sont employés dans les aéroports, gares, etc. où des caméras sont installées pour le contrôle de quais, aires clients, etc. La prise en charge de ces détections par des systèmes autonomes devient de plus en plus nécessaire devant l'accroissement significatif du nombre de lieux à contrôler, et l'importance également croissante de problèmes de sûreté publique. Un système d'analyse automatique de vidéos a une structure modulaire. Le premier module est la segmentation de la scène, c'est-à-dire la séparation de l'image courante en deux ensembles : la partie statique de la scène et les objets en mouvement et d'intérêt pour l'application. Cette phase est appelée détection d'objets. Les objets en mouvement qui ont été remarqués dans l'image courante sont mis en relation avec les objets en mouvement dans l'image précédente afin de trouver les correspondances dans le but de déterminer leurs trajectoires (phase de suivi d'objets). Ces trajectoires constituent ensuite l'entrée de la troisième phase où l'on analyse les comportements dans le but de déterminer les évènements de la scène. Dans cette thèse, nous proposons un système de vidéo surveillance qui, pour chacune des phases que nous avons introduites, présente des innovations par rapport à l'état de l'art, afin de palier les principaux problèmes qui se présentent dans le développement de tels systèmes. D'abord la plupart des problèmes relatifs à la détection d'objets ont été abordés. Il a été proposé un nouvel algorithme de soustraction du fond, sélectif et adaptatif, pour adapter le système à des changements de luminosité et de la scène. En outre, pour rendre applicable le système à des environnements réels, des heuristiques ont été proposées pour la résolution des différents problèmes : ombres, bruit, etc. Les résultats produits sur la phase de détection d'objets montrent que les techniques proposées sont robustes et utilisables en temps réels grâce à une temps de calcul peu élevé. L'objet principal de la thèse a concerné la phase de suivi d'objets. La problématique principale dans le domaine du suivi d'objets est celle des occlusions. On a une occlusion lorsque un objet est couvert par un élément statique de la scène (par exemple un arbre ou un mur) ou un autre objet en mouvement. Dans cette thèse, nous proposons un nouvel algorithme basé sur une représentation des objets basée sur les pyramides de graphes. Cette représentation permet de résoudre les occlusions, même dans les cas les plus complexes, en examinant l'objet à résolution croissante jusqu'à rejoindre le niveau de la pyramide où il est possible séparer les objets qui composent l'occlusion. L'algorithme de correspondance entre deux niveaux de la pyramide, pour relever l'identité des objets à partir des objets de l'image précédent, se produit dans un cadre basé sur les graphes en exploitant les relations spatiales entre les régions en mouvement. Des tests expérimentaux les relations spatiales entre les régions en mouvement. Des tests expérimentaux sur des bases de données standard et sur des index attestés pour l'évaluation des algorithmes de suivi d'objets en présence d'occlusions montrent que cette approche est très prometteuse. Mots clés : Analyse de vidéos-suivi d'objets-vidéosurveillance-détection d'objets-graphe pyramidale Détection, suivi, et analyse de comportement des personnes en mouvement dans les systèmes de vidéo surveillance : une approche basée sur les graphes Résumé In the last two decades we have been witness of a really grow up in the availability of audio-video information in several environment. Consequently, a remarkable interest has been recorded, in the scientific community of artificial vision. One of the most important domain has been the real-time video analysis, for the detection of events in video sequences. The detection of events in a video has several applicative implications. One of the most interesting application is that of the video surveillance systems. Video surveillance systems are employed in airports, etc. where cameras are installed are installed to control docks, areas customers, etc. These systems support the security operator, signalling anomalous events like as unattended luggage, or interdicted areas access, etc. This support is becoming more ande more because of the remarckable increased of the problems related to the public security. A video analysis system has a modular structure. The first step of the processing is the segmentation of the scene, that is the separation of the frame in two set: the static part of the scene (background) and the set of the moving objects of interests. This phase is called object detection. Detected moving objects are compared with moving objects in the previous frame to find correspondences. The aim of this step, called object tracking, is the objects along the sequence to determine the trajectories of the objects. Object trajectories are the input of the third step, behavioural analysis phase, which aim is to determine events that occur in the scene. Such events are the output of the system. In this thesis a video surveillance system is proposed. For each step it presents some innovations as regard as the state of the art in such systems. First of all, most of problems in object detection are tackled. A new, selectively and adaptatively, background subtraction algorithm has been proposed to adapt the system at illumination and scene changes. Furthermore, some heuristics are proposed to solve detection problems in real environment: shadows, noise, etc. Result show that proposed techniques are robust in terms of quality of solution and, besides, they are efficient in terms of processing time. The aim subject of the thesis concerns the object tracking phase. One of the most difficult problems in motion analysis is the occlusion problem. An occlusion occurs when an object is covered either by a static element of the scene (e.g. a tree or a wall) or by another moving object. In this thesis we propose a new algorithm based on a new representation of the objects: the graph pyramids. This representation allows the resolutions of occlusions also in complex cases. The objects are examined at different, increasing, resolutions until it is possible to separate the multiple region in the objects belonging to the occlusion. Furthermore, to detect the identities of the objects in the current frame beginning from the identities of the objects in the previous frame, a graph matching algorithm is performed. This matching algorithm exploits the spatial relationships between regions and it overcomes the limits of the algorithms present by now in the literature that are pixel-based. Finally experimental tests are performed. The are performed on standard datasets and standard indexes to provide objective results. The results show that the approach is promising.

IEEE Transactions on Systems, Man, and Cybernetics, 2004

Visual surveillance in dynamic scenes, especially for humans and vehicles, is currently one of the most active research topics in computer vision. It has a wide spectrum of promising applications, including access control in special areas, human identification at a distance, crowd flux statistics and congestion analysis, detection of anomalous behaviors, and interactive surveillance using multiple cameras, etc. In general, the processing framework of visual surveillance in dynamic scenes includes the following stages: modeling of environments, detection of motion, classification of moving objects, tracking, understanding and description of behaviors, human identification, and fusion of data from multiple cameras. We review recent developments and general strategies of all these stages. Finally, we analyze possible research directions, e.g., occlusion handling, a combination of twoand three-dimensional tracking, a combination of motion analysis and biometrics, anomaly detection and behavior prediction, content-based retrieval of surveillance videos, behavior understanding and natural language description, fusion of information from multiple sensors, and remote surveillance.

IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000

AbstractÐ R is a real time visual surveillance system for detecting and tracking multiple people and monitoring their activities in an outdoor environment. It operates on monocular gray-scale video imagery, or on video imagery from an infrared camera. R employs a combination of shape analysis and tracking to locate people and their parts (head, hands, feet, torso) and to create models of people's appearance so that they can be tracked through interactions such as occlusions. It can determine whether a foreground region contains multiple people and can segment the region into its constituent people and track them. R can also determine whether people are carrying objects, and can segment objects from their silhouettes, and construct appearance models for them so they can be identified in subsequent frames. R can recognize events between people and objects, such as depositing an object, exchanging bags, or removing an object. It runs at 25 Hz for 320Â240 resolution images on a 400 Mhz dual-Pentium II PC.

2008 19th International Conference on Pattern Recognition, 2008

This paper describes a visual surveillance system for remote monitoring of unattended environments. For the purpose of efficiently tracking multiple people in the presence of occlusions, we propose: (i) to combine blob matching with particle filtering, and (ii) to augment these tracking algorithms with a novel colour appearance model. The proposed system efficiently counteracts the shortcomings of the two algorithms by switching from one to the other during occlusions. Results on public datasets as well as real surveillance videos from a metropolitan railway station demonstrate the efficacy of the proposed system.

IERI Procedia, 2013

IOSR Journal of Computer Engineering, 2012

In current era of digital technology visual surveillance systems are persistently in pursuance of being easier to use, versatile, inexpensive and very fast. Continuous video capturing systems are the replacement for human watch, because as we know human can be easily distracted and one mistake may lead to big disaster. So video surveillance systems make this kind of work very easier for user and it provides security and control where all time watch is required. Proposed algorithm will helpful for to detect moving object and classify it as human being and keep track of moving human. This procedure is done without getting help of any additional sensing device. In this paper proposed system can classify in three steps detection, tracking and action analysis. Detection of human being is done by combination of morphological procedure and feature extraction method. Tracking of same human and occlusion handling is done in second phase. At last phase activity analysis is done and in case of any abnormal activities, an alert should be issued.

2013 IEEE Second International Conference on Image Information Processing (ICIIP-2013), 2013

This paper focuses on algorithms which are used to count the number of people moving in or out of an area supervised by a single fixed overhead camera. The algorithms presented here have the capability of determining people count for a single person as well as for multiple people crossing the range of camera. The overall mechanism has been divided into five modules and each one of them has been explained in detail.An efficient algorithm has been proposed for tracking single as well as multiple persons in the scene with the help of tracking using the center of gravity approach. Counting is basically done by tracking the person/people in the range of camera. The proposed system, however, faces certain limitations like the background must be constant, illumination should be invariant, static object problem etc.

In this paper, we propose a real-time video-surveillance system for image sequences acquired by a moving camera. The system is able to compensate the background motion and to detect mobile objects in the scene. Background compensation is obtained by assuming a simple translation of the whole background from the previous to the actual frame. Dominant translation is computed on the basis of the tracker proposed by Shi-Tomasi and Tomasi-Kanade. Features to be tracked are selected according to a new intrinsic optimality criterion. Badly tracked features are rejected on the basis of a statistical test. The current frame and the related background, after compensation, are processed by a change detection method in order to obtain a binary image of moving points.Results are presented in the contest of a visual-based system for outdoor environments.

IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000

is a real time visual surveillance system for detecting and tracking multiple people and monitoring their activities in an outdoor environment. It operates on monocular gray-scale video imagery, or on video imagery from an infrared camera. R employs a combination of shape analysis and tracking to locate people and their parts (head, hands, feet, torso) and to create models of people's appearance so that they can be tracked through interactions such as occlusions. It can determine whether a foreground region contains multiple people and can segment the region into its constituent people and track them. R can also determine whether people are carrying objects, and can segment objects from their silhouettes, and construct appearance models for them so they can be identified in subsequent frames. R can recognize events between people and objects, such as depositing an object, exchanging bags, or removing an object. It runs at 25 Hz for 320Â240 resolution images on a 400 Mhz dual-Pentium II PC.