A typical story: an incident happens, the archive is opened, and the image is so blurry that a face or a license plate cannot be made out. Or the archive ran out after three days when a week was needed. Both problems are built in at the design stage.
Start with a task for each camera
Before choosing a camera, decide what it needs to show:
- Observation: the general scene, movement of people and vehicles.
- Detection: noticing a person or a vehicle.
- Recognition: confidently recognizing a known person.
- Identification: establishing a person's identity beyond doubt or reading a license plate.
Each level needs its own pixel density on the target: the higher the level, the more pixels per meter of scene there must be where a person passes. The method for defining these levels is described in DSTU EN 62676-4. The required density determines the camera resolution, the lens field of view and the mounting distance.
A common mistake is installing one wide-angle, high-resolution camera "to cover everything". For identification at an entrance, a separate camera with a narrow field of view is better.
What determines archive size
The amount of video a camera records is determined by its bitrate, the amount of data per second. It depends on:
- resolution;
- frame rate;
- codec (H.265 usually compresses more efficiently than H.264);
- scene complexity (trees in the wind, rain and night-time noise increase bitrate);
- quality settings and recording mode (continuous or event-based).
How to calculate storage
Disk capacity for one camera:
Capacity (GB) ≈ bitrate (Mbit/s) × 0.45 × recording hours per day × number of days
(1 Mbit/s for an hour is about 0.45 GB.)
Example: a camera with an average bitrate of 4 Mbit/s, recording around the clock, 30-day archive: 4 × 0.45 × 24 × 30 ≈ 1,300 GB, i.e. about 1.3 TB per camera. Sixteen such cameras need about 21 TB of usable capacity, plus headroom and redundancy (RAID).
Motion-triggered recording can reduce the volume several times over, but for critical areas continuous recording at a reduced bitrate is better.
Recording reliability
- RAID arrays, so that a disk failure does not destroy the archive;
- UPS for servers and switches;
- health monitoring of disks and cameras with notifications;
- for critical sites, a standby server or duplicate recording.
Cybersecurity
Cameras are network devices, and they get attacked. The minimum set of measures: a separate network for cameras, changing default passwords, firmware updates, never exposing cameras directly to the internet, and secure remote access.
How we work
We define the task for each camera position, size the cameras and storage, design the network, install the system and configure analytics. More details on the "Video Surveillance" page.
