Abstract The article discusses the development of a system for recognizing people at a pedestrian crossing. Such recognition system includes a trained classifier and two sets of images taken from an open database containing images of city streets from outdoor cameras. Some shortcomings that can be easily solved with the help of the developed recognition system are analyzed. We also studied common methods of detection, for example, the Viola-Jones method, and their advantages in the operation of the system. A comparison of all the considered methods is carried out and the most suitable one for the set goal is selected and the method of constructing a classifier is described separately. This system is becoming more and more relevant with the development of the technological capabilities of urban life and improves its quality due to its smooth operation and good performance indicators.