The demand for surveillance systems has increased extremely over recent times.
Typical CCTV surveillance systems today:
- Raw live video stream of hundreds of (IP) cameras displayed on monitor wall
- Tracking: manually follow persons within a certain camera and in between camera views.
- Counter intuitive and fatiguing visualization.
- A more intuitive 3D visualization where the person’s path is integrated, independent of all cameras, is not yet available.
- Idea: Global tracking based on smart cameras and embedding results in one consistent 3D model
instead consists of a distributed network of smart cameras that allows for tracking and handover of multiple persons in real time and serves as basis for distributed activity recognition. The inter-camera tracking results are embedded in one consistent georeferenced 3D world model that is
accessible ubiquitously and can be viewed from arbitrary perspectives independent of the persons’ movements.
system comprises the following major components:
- A distributed network of smart cameras capable of embedded information processing in real time:
- Embedded object/person tracking.
- Embedded machine learning, e.g., activity recognition and object classification.
- A novel visualization within a consistent and geo referenced world model that is available ubiquitously.
We think it is more natural to compute the information where it becomes available - directly at the sensor - and transmit only results that are on a higher level of abstraction.
This represents the emerging trend of self contained and networking capable Smart Cameras.[more...]
Tracking plays a central role in many applications. Our system is capable of tracking multiple persons concurrently. It is very robust as it can handle multiple hypotheses and nonlinear motion patterns in parallel. Due to the SmartSurv
approach, persons/objects can be tracked throughout a whole network of camera nodes by automated handover.[more...]
system uses embedded machine learning techniques for various application fields, such as:
- activity recognition
- situation awareness
- object classification
offers a novel visualization that integrates all relevant results of the whole camera network in one consistent and geo referenced world model. It is available ubiquitously to an arbitrary number of users concurrently and shows all the tracking and activity recognition events as they happen.[more...]