Information technologies for video object recognition: monograph

The monograph is devoted to developing methods and means of using AI to rapidly identify moving objects in a video data stream based on deep learning technologies. Classical and non-classical methods of AI, convolutional neural networks, computer vision and pattern recognition, and theories of control systems based on estimates and criteria of mathematical statistics are considered. As a result of recognition, we will determine the type of the recognized object and will have quantitative accuracy estimates. A method for applying templates has been implemented. The algorithm has information about what the required object looks like, what kind of background it may have, how specific contours of the object look and what positions they may be. It immediately considers a possible location of the object's detection. It allows you to achieve high recognition quality and has good performance. However, when the video camera captures several similar objects, different patterns are satisfied, and recognition decreases. A family of models (artificial neural networks) are used to estimate or approximate functions that may depend on many inputs and are usually unknown.