Video analytics for a museum: flow management, security, and exhibition analytics
A large museum in Peru with high visitor traffic and a complex infrastructure reached out with a request: how to obtain qualitatively new information about what is happening inside and outside the building without changing the existing camera park?
Project background
A large museum in Peru with high visitor traffic and a complex infrastructure reached out with a request: how to obtain qualitatively new information about what is happening inside and outside the building without changing the existing camera park? The client was interested in objective data on the distribution of visitor flows across halls, automated security control (including night shifts), and analytics of audience interaction with the exhibition. A separate concern for the museum administration was whether the exhibits were placed correctly and how effectively visitor routes were structured. Heat maps revealed which exhibits remained undeservedly unnoticed and which attracted crowds, highlighting the need to review the logic of placement and exhibition rotation.
A solution was required that would allow the museum to shift from reactive management to proactive, receiving data in real time and making strategic decisions based on objective analytics.
Implementation process
A pilot solution based on the OSMOCV platform, using computer vision algorithms, was developed and presented. As part of the pilot stand, a connection to the museum's existing video streams was established.
Video streams from cameras installed in key zones were fed into the platform: the entrance group, thematic halls, perimeter, and parking lot.
The pilot solution included the following analytics scenarios:
- Automatic throughput counting — detection of incoming and outgoing visitors in the lobby and thematic halls for load analysis;
- Night monitoring and patrol control — recording the presence of people in premises after the museum closes, as well as tracking the routes of night security patrols (including arrival times at checkpoints and compliance with regulations);
- Behavior analysis — detection of falls, sudden movements, crowds, and conflict situations (fights);
- Exhibit protection — recognition of attempts to dismantle exhibits or excessive proximity to display cases;
- Route map construction (Heatmap) — visualization of “hot spots” to determine the most and least popular exhibits;
- Perimeter security — detection of open weapons, smoke, open fire, and low lighting;
- Parking analytics — automatic reading of license plates, recognition of car make and color;
- Visitor profiling — determination of gender and age (provided faces are visible).
Additionally implemented:
- automatic generation of alarm events upon violations with notification sent to the security service;
- generation of reports on attendance, exhibit popularity, and security service efficiency.
Achieved results
- Increased security level — the system allowed a shift from reactive mode (responding to the fact) to proactive (receiving a signal at the moment of threat). This is especially relevant for night hours, where additional control of guard patrols improves discipline and eliminates human error.
- Optimization of security service operations — security staff no longer monitor all screens simultaneously; attention is focused only on events highlighted by the system (alarm tags).
- Data-driven exhibition management — heat maps (Heatmap) revealed non-obvious patterns: a number of key exhibits turned out to be undeservedly neglected, while secondary items attract the audience. The data obtained highlighted the need to review the logic of exhibit placement and exhibition rotation.
- Understanding the visitor profile — collecting de-identified information on gender, age, and transportation preferences provided the administration with material for developing new loyalty programs and event activities.
- The ability to operate on existing infrastructure has been confirmed — the pilot stand proved that the system is capable of solving the set tasks without interfering with the client's physical infrastructure.