Feasibility Study of Artificial Intelligence Technology for Home Video Surveillance System

Currently, the concept of video surveillance at home can already be supported by a CCTV system or IP camera installed at home connected to an internet connection. However, the CCTV or IP Camera device is only equipped with minimal features, generally only a motion detection feature that will provide notifications or alerts if motion or motion objects are detected in the observed area. It causes inconvenience to the user because too many alerts appear if the area being observed is often accompanied by the movement of people or objects, which is a natural thing and does not need to be suspected. The application of artificial intelligence systems in the home area is expected to increase the capability of home video surveillance to be smarter in giving notifications/alerts. In this research, smart video surveillance is developed using two schemes of video stream processing: full local processing and semi-cloud processing. Two edge computing devices connected locally to the internet are installed in the home area. Both devices consist of Jetson Nano and PC GPU (Intel Core i7 7th Gen 7700HQ, 8GB RAM, Nvidia GeForce GTX 1060). The IP cameras used are 2 Ezviz brand IP cameras installed in two different locations and connected to one home internet network. Experiments show that the local processing scheme is much more recommended for a residential environment since it provides a much shorter response time than the semi-cloud one.