VIDEOSORVEGLIANZA Profili sociologici, criminologici, culturali e giuridici nell’ottica smart city Milano 24 Maggio, Università Cattolica del Sacro Cuore Videosorveglianza: la ricerca in Visione Artificiale per il riconoscimento automatico di persone ed eventi Prof. Rita Cucchiara, Università di Modena e Reggio Emilia, Direttore del centro Softech ICT http://www.Imagelab.ing.unimore.it Rita Cucchiara, Università di Modena e Reggio Emilia,
Welcome in Modena -University of Modena and Reggio Emilia The Second Short Spring School in Surveillance is organized by -University of Modena and Reggio Emilia -Softech-ICT Interdip. Research Center in ICT -Dipartimento di Ingegneria Enzo Ferrari - ImageLab Co-Sponsored by - Fondazione Cassa di Risparmio di Modena - GIRPR (Italian chapter of IAPR)
UNIMORE: Where we are… Università di Modena e Reggio Emilia Emilia Romagna, Italia Dipartimento di Ingegneria “Enzo Ferrari” ImageLab
WHY S5? A research School in (computer vision and pattern recognition for) Video Surveillance. Mainly a school for researchers few hot selected topics Short (but intensive) school in Spring period A Second edition…
S4 the 2011 edition Multicamera & applications Tracking Security and Forensics Face and expression Architecture and systems Large Crowd
Again Surveillance… Why another (short spring) School in surveillance? Surely is an industrial gold filed … is it still an open research problem?
The market The global Intelligent Video Surveillance (IVS) & Video Analytics (VA) industry revenues* totaled $13.5 billion in 2012, and are estimated to reach $39 billion in 2020. The rapid market growth is driven by the following dynamics: Increased use of video surveillance. Migration from analog to digital cameras and to IP based cameras. Technology maturity. Video analytics algorithms, processors, applications and products underwent a decade of technological evolution to intelligent video processing, based on advancements in image processing, enabling automatic detection of signatures detection and identification. Cost reduction of video analytic systems. Driven by the falling prices of image processing DSPs and communication systems. Cost-performance of new edge-based video analytics DSP technologies (e.g., Intel, Texas Instruments DSPs). From the "Intelligent Video Surveillance, VCA & Video Analytics: Technologies & Global Market – 2013-2020"
The market Human operators entail high cost & high rate of overlooked events. Real time analysis of video images and recorded footage is a need that can hardly be answered effectively by human operators, and manpower cost. Furthermore, human operators fatigue and boredom cause a high rate of overlooked events. It takes IVS, VCA and VA to process trillions of video surveillance-hours generated annually. By 2011 over 165 million video surveillance cameras installed worldwide captured 1.4 trillion video-hours. the investing growth of 9-11% year , captured video surveillance will reach approximately 3.3 Trillion video-hours in 2020. A hypothetical analysis assuming that 20% of the most critical video stream should have been reviewed by human operators, results in a (hypothetical) need to employ (by 2011) a workforce of over 100 million operators (working 8 hours daily, 300 days a year). Intelligent video surveillance and video analytics with mature technology have attracted a lot of attention. ---the global market* will triple by 2020.
THE RESEARCH 1998 the first workshop on AVSS (advanced video-based surveillance system) in Genoa 2000 The first IEEE Transaction PAMI Special issue on video surveillance 2003 the first IEEE conference AVSS ….. FIFTTEEN YEARS OF RESEARCH
Some statistics “Video Surveillance” : Google scholar 54000 items, IEEEXplore 6407 items from IEEEXplore papers with “video surveillance” keyword 2010 2011 2012 surveillance 863 867 768 analytics 23 27 46
statistics IEEE conference papers with video surveillance and ……
statistics IEEE journal papers with video surveillance and ……
surveillance 1) A growing interest in PEOPLE surveillance 2) From SECURITY to SOCIAL and WELLBEING 3) A research interest in computer vision AND learning 2010 2011 2012 traffic 117 104 86 people 83 85 93
2000 VideO Surveillance Research Acquisition Detection Tracking Single camera Multiple cameras VideO Surveillance Detection Moving object detection Single object/people Multiple targets, cluttered, occluded.. Tracking Linear prediction Generative models Stocastic prediction 2000 Research Understanding Purposive
PEOPLE Surveillance in the new decade Acquisition PEOPLE Surveillance in the new decade Detection & Tracking SPATIO-TEMPORAL REASONING Learning & Inferring BIG DATA ANALYSIS 2010 Research Searching & Recognizing HUMAN FEEDBACK
People Analysis IS PERVASIVE From Security Surveillance: Real time detection, Border, Zone control, Activity monitoring, Traffic, urban control… Safety: traffic-automotive control; crowd monitoring, home, working area control Forensics: Video Mining, Identity assessment, Likelihood ratio support To Multimedia Video analysis, broadcasting analysis, Digital Library annotation, Sport action recognition, Human Interaction, Gaming To (Social) Wellbeing Retail analysis, Comfort in architecture Design, Building Automation, Healthcare, Entertainment, Graphics..
People ANALYSIS Many applications.. Same problem: detect people, their identity and behavior
People surveillance: what at IMAGELAB Detection &Tracking People (and crowd) Detection and Tracking With multisensory data Learning& inferring People Identification and (Ri-)Identification Search & Recognition People Search by similarity Visual Learning A-priori Models
Surveillance at S5 The topics: Ci penso io quando avro’ gli abstract
SURVEILLANCE aT S5 The lecturers: Arnold Smeulders CWI Amsterdam, the Netherlands Bernhard Rinner Alpen-Adria-Universität Klagenfurt, Austria Francois Bremond INRIA, France Greg Mori Simon Fraser University, Canada
SURVEILLANCE aT S5 The lecturers: Horst Bischof Institute for Computer Graphics and Vision, Inffeldgasse Graz, Austria Larry Davis Institute for Advanced Computer Studies, Univ. of Maryland, USA Shmuel Peleg The Hebrew University of Jerusalem, Israel
SURVEILLANCE aT S5 The attendees
THANKS.. http://imagelab.ing.unimo.it Prof. Rita Cucchiara Dipartimento di Ingegneria Enzo Ferrari Res. Center SOFTECH-ICT Università degli Studi di Modena e Reggio Emilia Rita.cucchiara@unimore.it Imagelab Andrea Prati [IUAV-Venezia] Costantino Grana Roberto Vezzani Simone Calderara Giovanni Gualdi , Paolo Piccinini, Daniele Borghesani, Adnan Rashid, Paolo Santinelli Davide Baltieri , Dalia Coppi, Michele Fornaciari ,Marco Manfredi VISION-E srl Rudy Melli, Giuliano Pistoni, Emanuele Perini http://imagelab.ing.unimo.it