The latest image processing and video content analysis algorithms now enable the automatic detection of rule breaking actions and recognition of suspicious activities that make the work of security personnel more effective and more efficient. Possible functions that computers can overtake and solve in an automated way include
- tracking of persons or objects in one camera image or across multiple camera views,
- conducting identity tracking and remote personal identification,
- detecting entry into restricted zones,
- counting people or vehicles,
- detecting left behind or removed objects,
- a person fell, or someone lay down in an inappropriate location,
- recognizing other suspicious activities.
To realize these functions IDENTRACE solves the following basic video analysis steps:
- Background synthesis
- Static part of camera images, the background is determined and continuously maintained by the synthesis of the non-changing areas of video streams.
- Shape segmentation
- Shapes that move in the foreground are separated from the background by image processing algorithms.
- Object tracking
- Shapes are not only tracked within the view of one camera, but beyond that the objects are also followed on all camera images of the entire site, accomplishing real site-wide object tracking.
- Shape classification
- Segmented shapes can be classified by their different features, like size or shape. For example on a runway an airplane, a service vehicle, a person or even an animal can be distinguished.
- Feature extraction
- Properties of shapes - e.g. size or color - can be extracted from images. By focusing with a Pan-Tilt-Zoom (PTZ) camera on certain easily identifiable parts of shapes - e.g. faces, ears, number plates, bar codes - other, more detailed features, or even unique identifiers can be determined.
- Remote identification
- Based on certain features people or objects can be remotely identified. In some cases it is enough to know whether a person currently visible in a camera is or is not the same as somebody who was seen before somewhere else in the observed area.
- 3D localization
- With the help of the 3D world model created about the monitored site, the position of moving persons or vehicles can be precisely located.