The aim of the USafeS (“Urban Safe Services”) project is the digital transformation of safety in cities based on the fusion of data and big data.
The aim is to capture information from already-available systems, but additionally, the intention is to use other available non-structured data sources that are currently not used for the making of decisions in real time, such as data from video cameras (with their capture of video and audio), together with information coming from other channels. This requires heterogeneous real time listening systems, which must scale and deal with temporal peaks. This is why it is a challenge to combine techniques such as analysis of time series, deep neural networks and active learning and reinforcement systems, capable of learning which is the normal situation and capable of independently detecting when a behaviour is not normal, or is a natural evolution, or is a response to a specific event. Also, the aim is to reduce the existence of information silos. This implies sharing information between systems, and therefore it is necessary to anonymise the information, so that it can be used by third parties without any infringement of legal restrictions. To do so, it is necessary to have available a utilities and algorithms toolbox to anonymise data, so that if necessary, they can be traced, but without information which can be directly known.
Vicomtech’s role is focused on responding to the technological aspects most related to the computation of artificial vision algorithms and the management of video content from video surveillance systems for analysis, both in real time and offline.