The research in PAVIS follows two main streams, video surveillance and biomedical imaging/bioinformatics, being all centered on the analysis of data, images and video sequences in particular.

In this context, concerning the video surveillance area, we would like to explore how social signals can be utilized for the design of more effective surveillance systems, i.e., investigating how the findings derived from the human sciences (like psychology, anthropology, etc.) can be exploited to analyse the situations occurring in a scene, detecting, recognizing and possibly predicting abnormal behaviours.

To this end, we have to tackle in efficient ways typical issues like detection, tracking, recognition, in a word, investigating video analytics issues. More in detail, our aim is to study robust computational tools to obtain an accurate description of the dynamic scene viewed by a video camera or a network of cameras. Such algorithms have to be resilient to lighting variations, image occlusion and wide baseline views which are common aspects of video surveillance scenario. Besides, we will not only rely on standard vision sensors.

We aim to address the use of multiple and distributed sensors to achieve a more reliable monitoring. Three-dimensional (e.g., stereo) sensors, microphones, RFID, infrared/thermal cameras have all own peculiarities that can be exploited to obtain a more efficient supervision. Distributing such sensors in outdoor and indoor locations allows a complete coverage of the areas of interest, but also it requires an additional duty due to the need of communication among the devices. This leads to the necessity to process onboard most of the acquired data to limit the communication bandwidth. Hence, issues related to sensor networks, sensor fusion, and the design of embedded devices become other lines of research to be considered.

Reliable surveillance systems must cope, sooner or later, with biometrics issues, and in this context, we would like to consider non cooperative face recognition at distance, gait recognition, and re-identification.

The other main line of research is constituted by the analysis of images or, more in general, multidimensional data, which may derive from biological experiments (Genomics/Transcriptomics/Proteomics) and, more in general from biomedical instruments (e.g., MRI, PET, Expression Microarray, etc.) or from experiments carried out by the other IIT departments, especially Drug Discovery and Delivery (D3) and Neuroscience (NBT). Such kind of data may also derive from particular instrumentation, like electronic microscopes, so as to support Nanoscience facilities in their research.

In conclusion, the main research areas pursued in PAVIS are:

• Social Signal Processing
• Video analytics
• Sensors’ networks & sensors’ fusion
• Biometrics
• Biomedical Image Analysis & Bioinformatics
• Multidimensional Data Analysis & Applications