This paper proposes the possibility of recognizing pedestrians in real time on the end devices using a microcomputer to automate the process of signaling traffic lights and improve system mobility. The paper is devoted to a brief analysis of existing systems for the detection of pedestrians, with the subsequent development of its own system based on those studied, and its further implementation on the Raspberry Pi microcomputer. A natural experiment was conducted to detect pedestrians in an image with the subsequent determination of their location at a pedestrian crossing. A number of popular pedestrian detection systems were analyzed in terms of speed, accuracy of determination, and the possibility of implementation on microcomputers. Recommendations on the choice of the system, conditions of use and improvement of characteristics. The developed system automates the work of the traffic light through the introduction of additional functions (tracking a pedestrian's posture, moving pedestrians along the roadway), ensuring the safety of road users, which allows to improve the level of the urban environment as a whole.