Identification of People in the Video by the Characteristics of their Gait

Development object: an algorithm for identifying a person by the peculiarities of the gait and a software module that implements the developed algorithm. Objective: to develop an algorithm for identifying a person by the features of the gait and a software module that implements the developed algorithm and achieves an accuracy of at least 80% for a sample of at least 10 people. The work resulted in the development of an algorithm for identifying a person by the features of the gait, using as a sign of gait image of the energy of the gait, and a software module that implements the developed algorithm. The quality of the developed algorithm's work was evaluated in various experiments showing the influence of different viewing angles, wearing outerwear and backpacks or bags on the quality of the classifier's work, using different metrics (the percentage of correct responses of the algorithm; the percentage of objects called positive by the algorithm and thus really positive; the percentage of objects of positive class, which the algorithm has found, of all objects of positive class.). The results showed that the developed algorithm has achieved its goal and achieved more than 80% accuracy, namely 99% using the method of reference vectors on a sample of 50 people. The results obtained can be used to create software for video surveillance and perimeter security systems. To reduce the impact of covariate conditions on the work of the developed algorithm, it is recommended to extract the binary silhouettes without distorting them with outerwear, backpack or bag.