You are here

Low-Power Surveillance Camera

FBK ultra-low power cameras detecting the activity of the scene while overlooking the irrelevant parts

The IRIS Research Unit, in collaboration with TeV Research Unit (https://tev.fbk.eu/), has developed two sub-mW vision sensors, with VGA and QVGA resolution, embedding event detection capability. The two cameras are targeted to long-lasting battery-operated applications.

Connected devices are already populating our living environments (e.g., houses and work areas), improving our quality of life and our environmental footprints. While security systems, energy monitors and wearable devices are just few example applications which are already experiencing this paradigm shift, in the near future many more domains, such as transportation, smart buildings and robotics, are expected to observe the impact of the IoT wave.

Energy-efficient sensing, computing and actuating will necessitate a drastical re‑think of the system architecture and computing paradigm, aiming to achieve high performance and low power consumption at the same time. Among sensing technologies, vision is certainly that one requiring the largest amount of real-time computing resources. A hierarchical event-based design approach is therefore crucial to generate and propagate requests, from the lower levels (i.e., camera) up through the hierarchy to the more power-hungry processing components (detection and classification) and IoT layers. In this context, our vision sensors continuously process visual features on-chip at high efficiency to detect low‑level events aimed at triggering an external processor and deliver grey-level image and/or motion bitmap upon request.

This work was funded by the EU H2020-FCT-2014 FORENSOR Project under Grant n. 653355 (http://forensor-project.eu/).

Research topics: