What happened?
In the field of perimeter security, a new approach to video analytics is gaining traction: so-called cross-camera tracking. The company DEKI is developing a feature for its platform DEKI SAFE that evaluates observations from multiple cameras and identifies potentially related detections. The goal is to stop presenting individual detection events in isolation and instead display them in chronological order and in context.
The background to this is a fundamental problem in modern video surveillance: when multiple cameras in different locations on a site trigger alerts, security personnel often cannot immediately tell whether this is a single, evolving incident or independent activities by different people.
The details
Video analytics platforms such as DEKI SAFE already detect people in defined zones while taking predefined schedules into account. This makes it possible to identify activity in monitored areas outside permitted times. However, according to DEKI, a single detection provides only a limited insight into the overall situation at a facility.
The planned cross-camera tracking feature is intended to address exactly this issue: detections at different locations – for example at the outer fence and later at a restricted entrance – are to be presented as a possible sequence of a single event, rather than as two separate alerts. This would mean the security team no longer needs to independently compare footage from different cameras and manually correlate multiple alerts.
Technically, this kind of linking is demanding, especially when the system does not rely on biometric identification. Different camera angles, lighting conditions, distances, partial obstructions, and changes in a person's appearance all make visual comparisons more difficult. Two people wearing similar clothing may look alike, while the same person may appear quite different from another angle. This becomes particularly challenging when camera fields of view do not overlap – in that case, there is no continuous visual record of movement between two locations, and the system must assess whether separate observations could nonetheless be related.
DEKI emphasizes that a possible match between two detections cannot automatically be taken as proof that they involve the same person. For the technology to support operational decisions, security personnel must be able to understand which observations were linked, how confident the system is in that match, and which video sequences support it. Missing information and contradictory data must also be made visible. According to DEKI, the system's task is to support assessment – not to make the final decision.
Local processing and data protection
For operators of high-risk environments, such as critical infrastructure (KRITIS) facilities, the practical benefit of intelligent video analytics also depends on how and where data is processed. According to DEKI, DEKI SAFE performs video analysis locally on site using edge computing, meaning off-site cloud processing is not required. However, according to DEKI, local processing alone does not guarantee GDPR compliance. Particularly with cross-camera tracking, where observations from multiple cameras are combined into a more comprehensive picture of an incident, operators must retain control over who may access linked observations, how the information is used, and how long it is retained.
Context
The approach described marks a shift in perspective in video surveillance: away from the pure detection of individual objects or people, toward understanding an entire sequence of events. For security teams in perimeter protection applications, this represents a potential reduction in workload, since individual alerts no longer have to be reviewed in isolation. Instead, related sequences of events could be presented in a structured way.
At the same time, the source material makes clear that this technology is still in development and validation. As part of a project with European authorities, DEKI is developing and validating application scenarios for advanced surveillance and perimeter protection. This work is part of a broader initiative aimed at enabling context-based video analysis and earlier detection of security-relevant events. According to DEKI, the suitability of the approaches under investigation for routine use still needs to be demonstrated, as does the reliability, transparency, and practical benefit of the planned cross-camera tracking feature.
Practical tips
- Operators of security-critical facilities should check, when selecting video analytics solutions, whether and how observations from multiple cameras are placed into a traceable context – not just whether individual objects are detected.
- It is important that a system discloses the basis for its matches: which video sequences support a link, how confident is the match, and where are there gaps or contradictions?
- For systems with local processing (edge computing), operators should additionally establish their own rules regarding access rights, use, and retention periods for linked data in order to ensure GDPR compliance.
- According to DEKI, the technology should be understood as supporting assessment, not as an automatic decision-making authority – human review remains necessary.
Outlook
In the medium term, a shift could take place for security personnel: away from reviewing individual alerts, toward examining a more comprehensive sequence of events. According to DEKI, whether cross-camera tracking can live up to this ambition in practical operation depends largely on how reliable the matches are, how transparently they are presented, and how well they integrate into the actual daily workflow of security teams. Further validation as part of the government project is expected to show whether the approach can move beyond the concept stage toward broader deployment in perimeter security.
