Artificial intelligence will bring real benefits there where it copes with some tasks better than a person. There are many such tasks in video surveillance.There are many such tasks in video surveillance. Firstly, the operator is not so good at viewing the image from many cameras in real-time, as is expected when designing control centers. Secondly, the volume of saved videos is now incredible.
At this stage of its development, artificial intelligence is built on the principle of a neural network. It is similar to the human brain, in which bioelectric signals are transmitted along neurons. Neurons are interconnected at nodes called synapses. To prepare the network for work, it is necessary to conduct its training, during which a certain connection between the input and output signals will be established in each node.
In it, the signal filtering in the synapse is performed according to a certain mathematical algorithm. The training of such a network is slow, it needs to present thousands, perhaps millions of examples, marked as “necessary” and “not needed”. The learning process may include thousands or millions of cycles. Determining which impulse made the greatest contribution to the overall result, it increases its contribution to subsequent summations. Such a mechanism gives a result much faster and, according to a number of experts, better copies the work of the human brain.
The competition between these types of networks, which some new ones want to join, is still ahead. And since the scope of commercial products with artificial intelligence will not only be video surveillance, it will also matter how a certain type of artificial intelligence worked out in one area can be used in others.