Video Patrols with AI support

Optimize customer security costs without reducing your security level by implementing AI Video Patrols based on Safestar

Contact with a specialist

How AI supports video patrols in Safestar?

AI plays the role of an artificial operator performing the same activities as natural operators.

The artificial operator independently handles video patrol, evaluates images according to given criteria and, depending on the situation, either closes the event or, in the event of detecting a threat or irregularity, directs it to further handling by a human operator.

Overview

Video patrol in Safestar is an optimized video monitoring with extremely high efficiency. It reduces the number of guards on the protected object without compromising its security.

Its operation is based on algorithms that allow the presentation of specific images at a particular time. You can flexibly configure the video patrol and decide when the natural or artificial operator will see a given image and detailed information.

In addition, guidelines for control and observation can be presented in a selected form – reference frames or described control procedures.

Why is it worth using video patrol in Safestar?

You get a complete and ready-to-use tool. You can create new, more optimized, attractive protection services for your clients. You will make your new image as a professional protection company that uses AI-supported modern technologies.

Benefits

It’s not enough to say that AI reduces costs. See what exactly it is and what benefits it brings you.

Automatic video patrol handling by AI

Safestar automatically connects to cameras on protected sites and performs AI comparative analysis to detect significant differences in views. Only when irregularities are detected, Safestar forwards the video patrol to the operator for handling. When Safestar AI does not detect anything disturbing, the video patrol event is closed with an appropriate description. It is worth noting that when connecting to the cameras, the video stream can be automatically recorded in Safestar for later verification of the quality of the service provided.

Patrols schedules with random option

For each video patrol, an advanced work schedule can be created, also taking into account dynamically changing non-working days and holidays. This means that VP can be perfectly adapted to the work schedule of the protected facility, which allows you to avoid unnecessary patrols. The exact time of the patrol can be randomly modified, so as to prevent the risk of deceiving the patrol.

Automatic photo-video documentation

Automatic creation of documentation from video patrol allows for later generation of reports from the inspection of the facility containing both photo-video material and additional information regarding the course of the patrol. Thanks to this, managers of security companies and owners of protected facilities can receive detailed information in the form of reports sent by e-mail.

Reference frames and parameter customization

Regardless of the situation, operators can record images from cameras in real time during a video patrol. These images can be crucial in documenting various events, such as suspicious behavior or a security breach at a monitored facility. It is also worth noting the wide possibilities of customizing video patrol parameters for both manual and AI-based patrols.

How does AI Video Patrol work?

AI video stream quality tests

Video stream quality tests are a key feature that allows you to always maintain the correct quality of video surveillance.

Safestar AI automatically checks the image quality from the monitored cameras at specified intervals. The check includes, among others, the absence of the image, sharpness, whether the image is too bright or too dark, and challenging weather conditions that prevent proper observation. If the AI assesses that the video from the camera is of insufficient quality, it will alert the operator by generating an appropriate event.

Camera image freezing

A common problem that is difficult to detect is the “freezing of the image” from the camera. The video stream seemingly exists, so the analytics can analyze it and the naked eye will not catch the irregularities without careful attention. Safestar AI does this automatically at user-defined intervals and standardly during video patrol.

Poor lighting conditions

In many regions, we are dealing with large changes in lighting conditions. Even with well-chosen cameras, frames and camera locations, underexposure and overexposure, fog, heavy rain and smoke can occur. In such conditions, video analytics may not detect threats correctly. Safestar AI detects such situations and informs operators about them.

Insects, spider’s, etc.

In outdoor installations cameras are exposed to the impact of various insects. They can effectively cover a large part of the image or make it unusable. IR lighting further aggravates these circumstances, when even the smallest threads of spider’s web can ruin the image at night. Our AI also detects these anomalies and informs the operators about it.

Frequently asked questions:

We kindly answered any questions regarding AI Video Patrols in Safestar

What are AI-supported video patrols?

With AI support, we can automatically check whether the live image is similar to the image from the previous patrol. In the configuration, we specify the minimum degree of similarity required to the last image. AI analyzes the image and, based on advanced algorithms, checks and decides whether such an image should be forwarded to the operator for verification.

Can we set the degree of similarity for each camera?

Yes, when defining a video patrol, we collectively define this degree for all cameras used in the patrol. However, if we want to change it for a single camera, we change it in the stream configuration and this will be the superior setting to the general one.

What are the degrees of similarity for video patrol?

There are three degrees of similarity for a video patrol – high, medium and low. High equals 98%, medium 95% low 90%. If the similarity to the previous patrol is lower than the set threshold, such an event will be directed to manual operation by the operator.

How does the operator receive information about manually checking?

When the operator should handle it, the event is shown with the type and priority set in the patrol definition. Additionally, during handling, the operator is shown in red cameras with a similarity lower than the set threshold and information about how much the image differs from the previous one.

What happens in the case of positive verification of all cameras?

Then, the system will automatically end such an event without involving the operator. This will be noted in the object history and the images from the cameras during their analysis.