So far, the majority of systems to monitor and analyse the behaviour of crowds were based on analysis of data from surveillance cameras. These approaches present some drawbacks and weaknesses due to factors such as spatial limitation of the cameras coverage area, occlusion, illumination conditions, etc. Recently, smartphones and wearable sensors began to be used to observe and analyse social interactions and behavioural dynamics. The richness and accuracy of data collected by the sensors embedded in the smartphones make possible to monitor and observe a wide range of daily activities of human beings and recognize the contexts of use.
- Their coverage area is unlimited (however, GPS does not work indoor).
- They do not suffer from occlusion problems
- They systems do not depend on lighting conditions.
- They allow collecting data on a large scale (therefore, they are very suitable for monitoring major events).
- The sensors are already present in the smartphone without requiring addition sensors.
The data so collected will be processed by machine learning algorithms to recognize groups, their behaviour and therefore predicts and detects potentially dangerous situations to make public event and our cities safer for everyone.
Furthermore, the use of data collected from the mobile devices requires specific solutions to address the problem of secure data sharing and confidentiality. As part of this project, we will study how to secure and make anonymous individual data (protection of personal and identifiable information) and collective data (protection against sensitive learning knowledge representing the activities of a group) collected with phones mobile while maximizing the efficiency and accuracy of machine learning algorithms.