UNIVIEW AI BOX

Add Cameras to the AI BOX

The process begins by adding cameras, either via ONVIF or RTSP. These can be incorporated manually or through a search on the local network.

Algorithm Capability Config

In this module, detection algorithms are managed, allowing them to be enabled or disabled so they are available in the cameras' analytics configuration. The system allows up to 20 algorithms to be selected.

Different predefined scenarios are also shown. When one is selected, the recommended algorithms are automatically adjusted according to the type of installation (e.g., hospitals or schools). However, any algorithm can be manually selected according to the project's needs.

Smart Config

In smart config, the previously enabled algorithms can be added to each of the cameras.

These parameters can also be configured directly from each camera, within the device addition section.

Resource Management and Selection of Analytics Rules

Within the analytics configuration section, available rules can be selected, taking into account the percentage of assigned resources. When this percentage reaches 0%, it is not possible to add more rules to the different cameras.

Dress Code Control and Verification of Authorized Profiles

In the dress code section, a list is provided to verify whether personnel in a facility comply with the mandatory use of work clothing. In addition, the system allows the registration of relevant profiles for security purposes, in order to quickly identify individuals whose access or presence needs to be supervised according to the facility's operational criteria.

Event Search and Facial Match Management

In this section, the system allows you to consult scheduled events and, if facial detection is enabled, to perform face searches to incorporate relevant matches into the profile library.

System Optimization through Self-Learning

Self-learning, when enabled, allows the device to continuously refine its analysis capability according to the configured rules. In this way, the system improves its detection by recognizing previously identified behaviors or elements, increasing effectiveness in future alerts.

Example: When the knife detection algorithm is activated, the system initially identifies the person carrying the object. If self-learning is enabled, the device associates that pattern with the individual and, in later appearances, can continue classifying them as a previously detected subject, even if the object is no longer visible.
In the first image, it can be seen that the individual is carrying an object identified by the system as a knife. The analytics module detects the corresponding pattern and highlights the subject with a box, automatically generating the event associated with that detection.

In the second image, it can be seen that the individual is no longer carrying the object or it is not visible. However, thanks to the self-learning process, the system recognizes them as a subject previously associated with a knife detection, maintaining the classification based on the analytics history.

Real-Time Monitoring via Data Panel

The system incorporates a live data panel that allows security personnel to monitor generated alarms in real time, improving immediate response capability.

IMPORTANT: To ensure accurate detection, the system requires that subjects are preferably captured from the front. In addition, cameras should not be installed at excessive heights, especially when it is necessary to identify small objects or perform facial recognition, as a high perspective reduces the sharpness and effectiveness of the analysis.