Visual surveillance systems for the identification and characterization of anomalous behavior



Current work in the field of security and surveillance (as a result of advances in the development of low cost processors, high performance and mobile communications systems of information) are directed to the development of distributed computing systems based multimedia multimodal data entry and have a capacity intelligent understanding of the scene focused on the automatic generation of alarms or focus of the operator. The distribution of intelligence is a crucial factor of these systems. Different devices involved in the work of surveillance (cameras, hubs, control centers, ...) (Foresti et al., 2003) are able to perform logical operations and, therefore, to pass or consider only the information of interest. This causes a considerable improvement in performance by reducing processing time, the required bandwidth and computational load of each of the elements. They can also take advantage of the evolution in the wired and wireless transmission of multimedia information. For all these reasons, research in cooperating devices integration of perception and action and between them and human activities, and computer network management, both corporate and specific results to such integration; distributed processing techniques in real time video, computer vision, virtual reality-based care events, scene analysis, digital communications, robotics, systems, ... they predict interesting results directly applicable video system (Hampapur et al., 2005).

From a technological point of view, a surveillance and security system will consist of (Velastín and Remagnino, 2006):

  • System Architecture: Infrastructure communications, sensing, control and performance.
  • Operative system: integrating software element of the different features provided by the system. To isolate applications from the specifics of the technology used, separating these two layers, this independence facilitates the upgrading and maintenance of both technological and application-level .
  • Information system: annotated video storage and configuration of security and surveillance system.
  • Services: system monitoring functionality. This includes tasks such as motion detection, tracking, person identification, interpretation of movement or access control. Many of these tasks are common to different actions to maintain the integrity of the facility and property. For example, motion detection tasks can be part of intrusion detection, perception, attack prevention and others. You must also provide services for the subsequent revision of the videos. Must also appear qualified services to perform management tasks and system maintenance.

The surveillance process with a single camera includes different phases: modeling of environments, motion detection, classification of moving objects, tracking, analysis and description of behaviors, identifying people ... With the objective of extending the monitoring to a global area and, under certain circumstances, to avoid occlusions, it is necessary to merge data from multiple cameras (Hu et al., 2004).

In recent years, has worked successfully in the analysis of recent performance by effective and robust techniques for detecting and tracking objects and people that have allowed the shift of interest towards higher levels of understanding of the scene. Moreover, thanks to the proliferation of cheap sensors for vision, embedded processors, and efficiency of wireless networks, much research has focused on the use of multiple sources of information for the analysis of behavior. It has also developed potential new applications such as human-computer interfaces for virtual reality, games, gesture-based control, presence and events for ambient intelligence applications or human-centric applications such as detection of falls in the eldercare. Besides the obvious advantages, the use of vision sensor networks incorporates several challenges such as management of data and data fusion, distributed processing, and algorithmic techniques for efficient extraction of interesting behaviors from redundant data

There are plenty of research on the use of recognition techniques postures and movement analysis. However, it appears that the integration of vision systems, security environments is still scarce.

This is the main question that project attempts to solve:

  • Analyze and develop proposals to integrate advances in machine vision that are susceptible be transferred to the world of visual surveillance.

Back to index


The overall objective of the research proposed in this project is the specification and design of vision-based services aimed at facilitating the monitoring of an area with poor visibility. The camera is one of the most versatile sensors. With it, you can develop many services to improve visibility in space with obstacles or poor lighting conditions.

It is proposed to specify and design a visual surveillance system in real time on graphics processing units (GPU) with capabilities to identify and characterize abnormal behavior of one or more individuals in restricted environments by representing and interpreting the movement in support of preventive actions security systems. The proposed system will also activate mechanisms that facilitate the analysis of situations (eg activate illumination of an area or direct the camera to a target). Besides the vast computing power of GPUs allow simultaneous treatment of sequences of images obtained from multiple cameras and active representation by models based on neural networks. Just as the classification and recognition of abnormal behaviors.

Also apply characterization techniques and recognition of movement to facilitate recognition and tracking people.

  1. Specification and design of an information processing system about human figures from images of low quality.
  2. Specification and design of an identification system to identify unusual situations risky behavior by analyzing image sequences from captured video.
  3. Development of a prototype. From vision application design a system architecture is proposed to develop a prototype on GPU to validate the system in a restricted environment such as the lobby of a building, a car park or camping.

Back to index

Background and previous results of the group

The research team of this project belongs to the Singular Research Unit "Industrial Computer and Computer Networks" Department of Computer Technology at the University of Alicante. The lines of work within the group integrate artificial intelligence techniques, control, vision and networks having different contributions made applicable for industries (machine vision systems, control of mobile robots, design and computer aided manufacturing), and human interaction machine (gesture recognition or representation and tracking people and objects).

For over ten years has been getting public funding for projects related to computer vision in different fields of application: man-machine interaction (CTIDIA/2002/109, GV06/195, FIT-350200-2005-54 ), robotics (MCYT-DPI200204434C0401, CICYT-TAP98-0333-C03-03, CICYT-TAP97-1164-C03-02, GV-CTIDIA/2002/183), virtual prototyping (DPI2005-09 215-C02-01) or visual inspection (MEC-1FD970614, CICYT-TAP96-0747, GV05/181).

The main researcher has developed his doctoral thesis on the development of models representing objects and their movement in realistic environments under time constraints. As applications of representation system are proposed a visual surveillance and human-machine interaction systems. Also have established stable cooperation with research groups in the UK by making stays of the researchers involved in the British universities of Westminster, Queen Mary and Kingston.

Back to index

Expected project benefits

By previous experience in this field and by publications and projects done in years past, we hope to make the following scientific contribution:

  • Preparation and presentation of the doctoral thesis by research fellows of the team members and doctoral associate professor.
  • Organizing an international workshop or conference related to the project theme
  • Presentation of papers at international conferences (ICCV, ECCV, CVPR, IJCNN, etc..) That will form part of the plan for dissemination among the scientific community.
  • Publication of articles in prestigious international journals: Neurocomputing, Pattern Recognition, Applied Soft Computing, Image Understanding and Vision, etc. to validate developed techniques.
  • Incorporation of project results to the Doctoral Program with quality mention "Information society technologies" Department of Computer Technology at the University of Alicante which won the quality award for 2008-09 to 2011-2012.
  • Creating a unique site which will detail the project and its results

Back to index