Wednesday, September 9
Smart image sensors : new paradigms or back to old concepts?
Antoine Dupret, CEA-LETI, Grenoble (France)
There was less than one decade between Carver Mead's electronic retina (the early 1980s') and Eric Fossum's Active Pixel CMOS imager. Within the next decade the CMOS image sensors have gained a dominant position in imager's market by replacing CCDs in almost any application. Since that time, the design of image sensor has been driven by the demand for, first of all, pixel density for Digital Still Camera, and secondly for better images. To keep on the race towards ever smaller pixels and with lower noise, many recent market-oriented works emphasis on new pixel structures or 3D IC stacking, even if recently new functionalities such as 3D image sensors emerge. On the opposite the concept of electronic retina seems to have failed to find a commercial success. Yet, one of the very first CMOS optoelectronics circuits was Lyon's optical mouse. Lyon pioneer's works has settled some of the fundamental aspects of design methodologies for smart image sensors, i.e. use of very simple, not to say conservative devices, mixing analog and digital circuits, facing the challenge of their implementation in the reduced area devoted to the processing elements. The conflicting requirements in image acquisition lead to explore novel optoelectronic sensors, and to revisit the concept of smart image sensor, for instance by considering the implementation of recent progresses in the domains of signal processing and of micro-optical devices.
Dr. Antoine Dupret graduated in electrical engineering from the Ecole Normale Supérieure de Cachan, France. He received the D.E.A. (Master degree), PhD, and the "Habilitation à Diriger les Recherches" degrees from the Université de Paris Sud 11 in 1991, 1995, and 2004 respectively, all three in electrical engineering. In 1996, he held an assistant professor position with the Université de Paris 13, and was researcher at Institut d'Electronique Fondamentale (IEF), Université de Paris Sud 11-CNRS, France. His works were about smart image sensors and compact model baseed sensor/processing co-design for MEMS. In 2009, he held a full professor position at ESIEE Engineering, in Noisy-le-Grand, France. He joined CEA-LETI in 2010 where he is senior expert in Image sensors. His current interests are in vision circuits (heterogeneous image sensors; silicon retinas) and opto-electronics. He has successfully leaded, and is currently leading, several projects related to smart image sensors and heterogeneous integrated systems design. His projects have been funded by French public research agencies (ANR, DGA...), and by industrial partners. He is author or co-author of over 80 papers in International Conferences and Journals. He served as a program chair of several international Workshops.
Thursday, September 10
Embedded Video and Fusion Analytics Solutions for a Global Intelligent Lighting Platform
Csaba Rekeczky, Eutecus Inc. Berkeley, CA (USA)
Eutecus, Inc. has been developing video analytics systems for many years. Based on our Multi-channel Video Analytics Engine (MVE™), these systems make use of real-time video-based algorithms that have the ability to understand essential information in the scene. Currently, we are incorporating such miniature cameras and credit card-sized video processors into Solid State Lighting (SSL) systems to support the Intelligent City initiative. Primary goals of these Intelligent Lighting Systems (ILS) are to provide intelligent controls that will enable significant reduction in our national use of electricity as well as a means to enhance global safety and security by providing non-conventional surveillance and filtered event-based information to law enforcement agencies and municipalities.
It is estimated that thirty billion lights exist around the world. Imagine light fixtures (i.e., luminaries) which not only maximize lighting efficiency locally but connect with other similar devices globally to make efficient use of energy and share information to those who use and need it. By 2050, fifty billion such connected devices (e.g., lights, refrigerators, thermostats, turbine engines, full cities) are expected to exist (source: Cisco, CCS 2013) and the penetration of such connected things is expected to grow from 0.6% in 2012 to 2.7% by 2020. When one considers the sheer volume of data produced by such devices, especially accounting for the rapid growth of high-definition video sensors, the numbers become staggering and it is clear that the Cloud will not have the required bandwidth to support this volume of data.
With their technology partners, Eutecus has realized that only the most relevant data be produced at the luminaire before being transmitted to the Cloud as a means to control bandwidth requirements; we have made significant progress to date and more is required. Also as a means to help safeguard our privacy, our products do not stream personal information to the Cloud but restrict such information to details of events (time, location, type of event, etc.) which have occurred and were detected by our video analytic methods. Typically, such information relates to counting vehicles, pedestrians, available parking spaces, traffic violations and other safety-related events.
When part of an ILS system, Eutecus' ReCo™ family of miniature video analytic camera/processor systems (ReCo™-Duo dual-camera and ReCo™-Pro four-camera) video analytic systems add significant value and functionality.
Dr. Csaba Rekeczky contributed over a decade to the research, innovation and design of massively parallel algorithms and cellular multi-core architectures for various application areas ranging from medical imaging to surveillance and reconnaissance, then went on to transition and commercialize technology achievements in start-up and spin-off company arrangements. He received the M.S. degree in electrical engineering from the Technical University of Budapest in 1993. After graduation he joined the Neuromorphic Information Technology interdisciplinary postgraduate program and continued his studies at the Analogical and Neural Computing Systems Laboratory of the Computer and Automation Institute of the Hungarian Academy of Sciences. As part of his PhD studies, in 1994 and 1995, he spent a year at the Tokushima University (Tokushima, Japan) as a visiting scholar working on research projects focusing on the application of cellular neural network based computing to medical image processing. In 1997 and 1998 he conducted research in nonlinear image processing and neuromorphic modeling of the vertebrate retina at the University of California at Berkeley (Berkeley, USA). He received the PhD degree in electrical engineering from the Budapest University of Technology and Economics in 1999, and spent a post-doctoral research period at UCB during 2000. Along with his co-authors, he won the Best Paper Award for a contribution published in International Journal of Circuit Theory and Its Applications in 2000. In 2001 and 2002 he served as one of the Associate Editors for IEEE TCAS-I. Several years later, he also served as an Associate Editor of the International Journal of Circuit Theory and Applications (2010). In 2000 he co-founded AnaLogic Computers Ltd. in the EU, then in 2003 Eutecus, Inc. in the US, and in recent years played a major role in transitioning a cellular nonlinear network based multi/many-core processor and an associated massively parallel algorithmic design technology from the academic environment to the industry. Currently he serves as CEO and President of Eutecus, Inc., a high-tech US Silicon Valley company with an EU subsidiary developing products and services at the fore front of multi-core video and fusion analytics technology. He is currently on leave from his Associate Professor position at PPCU, but maintaining a very active collaboration with and supervising several graduate students through the Virtual Video Analytics Laboratory, recently established by PPCU and Eutecus, to promote engineering research in the quickly emerging technical and business segment of Advanced Driver Assistance Systems.
Friday, September 11
Use of Distributed Smart Cameras for research in ecology: the case of the lesser kestrel
Javier Bustamante, Doñana Biological Station EBD-CSIC, Seville (Spain)
Much research in ecology is based on direct observation of animal, plants or natural processes. But observing nature is not easy. For example, animals are shy and the observer frequently interferes with the behaviour of the animal he tries to study. In this situation, cameras can be of great help. I will introduce a research project on an endangered species, the lesser kestrel (Falco naumanni), a small falcon of colonial habits, for which we have been using Information and Communication Technologies (ICT) for the long-term record of behavioural and population dynamics parameters at a breeding colony. The lesser kestrel is an insectivorous species breeding in colonies. It used to be very abundant in agricultural landscapes around the Mediterranean. In the 60's and 70's of the XXth century it suffered a marked decline associated with agricultural intensification and the abandonment of marginal agriculture. The species disappeared in many European countries and declined in Spain. Currently is considered an endangered species. At Doñana Biological Station we have been studying the ecology of the species since the 1980's. In 2006 we started a research project named HORUS with the aim of using sensor networks to study breeding behaviour and the long-term dynamics of the species at a breeding colony. We have developed a "smart nest-box" equipped with sensors that collects information on the kestrels breeding inside. We have implemented RFID (radio-frequency identification) tags that are read by an antenna when kestrels enter the nest-box, IR barriers that record movement direction, a balance that weights the individuals, temperature sensors, and video cameras. Each nest-box is controlled by an Arduino, and all nest-boxes are connected to a network and transmit data to a centralized database. A pilot system was installed at a building holding a grain elevator at La Palma del Condado, Huelva, Spain, where kestrels where nesting at the window sills. The kestrels accepted the smart nest-boxes hand have been using them to breed since 2009. The system is connected to the internet so it is possible have remote access to the cameras, the data that are collected, and to operate the system. The camera in the nest-box is programmed to record a a single frame and a 10s video sequence with movement detection and all videos and photos are stored on-site. Researchers can access to the live streaming of a single camera, to a mosaic of all cameras, to photographic summaries of images taken at specific times, or to all images recorded. I will show some of the results obtained in the project and discuss the new problems that working with cameras and images represent for us.
Dr. Javier Bustamante graduated and completed his PhD in Environmental Biology at the Universidad Autónoma de Madrid in 1988
and 1990, respectively. He held post-doctoral positions at the CSIRO Division of Wildlife and Ecology, in Canberra (Australia),
Berchtesgaden National Park (Germany) and the Doñana Biological Station, where he is a staff Research Scientist since 1996.
He is currently working in spatial ecology. His background is that of a field ornithologist, that has gradually moved more into mathematical models of species distributions, statistical analysis, geographical information systems and remote sensing.
He started working in behavioural ecology of birds of prey (Parent-offspring conflict in kites, ospreys and kestrels was the subjects of his PhD thesis work). He did some research in penguin reproductive ecology --Chinstrap penguins-- in Antarctica, and
then moved into species distribution modelling (SDM). The fact that environmental predictors in species distribution models
were needed forced him to start working with GIS and remote sensing. He continues working with Lesser kestrels (Falco naumanni) in a long term research project. They are building an automatic monitoring system for a breeding colony in Andalusia using GPS-dataloggers, accelerometers and smart nest-boxes. He also does research in species distribution models in Spain and South America.
He is interested in different aspects, but specially in building tools that can be used in species conservation, management, and conservation planning at different spatial scales. In remote sensing, he is particularly interested in the use of satellite images for time series analysis, to be able to reconstruct how natural systems work at large spatial scales. He currently works in a project on historical reconstruction of the flood dynamics of Doñana wetlands.
He has been member of the Board of Directors of the Society for Conservation Biology - Europe Section and
Associate Editor of the Journal of Applied Ecology.