Industrial Automatic Control Systems and Controllers Annotation << Back Mathematical Approach and Model for the Problem of Detection and Classification of Vehicles on Photographic Images of a Scene of a Street-road Network V.S. Tormozov, A.L. Zolkin, M.R. Bogdanov, S.V. Vygovskiy, T.G. Aygumov The article discusses a mathematical model and an algorithmic approach to the current task of detecting and localizing vehicles on a sequence of digital images of a scene of a transport section of the road network. The general principles of operation of the system for counting and classifying transport, using a mathematical model of the scene and the function of image intensity, are presented. This system is not only an efficient and cost-effective way of collecting traffic data, it also opens up new possibilities for research in the field of machine vision. The method of contour classification using mathematical operations of cluster analysis, proposed in this work, is unique in its way and opens another niche for research in the field of machine vision. The system works without reference to vehicle numbers: the vehicle is recognized even if its license plate is not visible. A modified vehicle detection algorithm is proposed, a feature of which is clustering of deterministic object descriptors. The main stages of vehicle detection are considered. The results of studies are presented, showing the fundamental possibility of using the proposed method for the practical task of collecting data on city traffic flows. Keywords: gradient descent method; function of several variables; digital image processing; contour analysis; search for the global minimum. DOI: 10.25791/asu.11.2021.1327 Pp. 31-38.