In this thesis, a locally-hosted video surveillance system was developed for a residential building. The fundamental element of the video surveillance system are IP cameras, which were selected on the basis of the OODPCVS standard. These are connected to a local network. The local server runs the Frigate platform, which uses computer vision and artificial intelligence to detect events in the camera footage and stores them on a local disk, whereby sensitive data never leaves the local network. In addition, the server performs automatic data backups and enables remote access via a VPN connection.
The system reliably detects people, animals, and vehicles under most weather and lighting conditions. Most of detection difficulties occur at the edges of the field of view, where objects are represented by a smaller number of pixels. It turned out that a local implementation of video surveillance with local server hardware makes the most sense specifically when one wants to run other services on it in addition to video surveillance, and when there is sufficient time and knowledge available to set up and maintain the system.
|
