This study deals with the problem of rea-time obtaining quality data on the road traffic parameters based on the surveillance camera data. The purpose of the paper is to develop a system to collect data on the traffic flow structure and to determine the traffic flow speed and direction in real-time. Our solution is based on the use of the YOLOv3 neural network architecture and open-source tracker SORT. To increase the accuracy of detection and classification, we used multi-scale prediction with an increased number of anchors. To determine the speed, we used a matrix of the perspective transformation of the source image to geographical coordinates. To train the neural network, we marked over 6,000 images and performed augmentation, which allowed us to increase the dataset to 60,000 images. Checking the system at night and in the day showed an absolute percentage accuracy of counting vehicles of no less than 92%. The error in determining the vehicle speed by the projection method, taking into account the camera calibration, did not exceed 2.74 m/s. The presented study allows us to generate big data for the intelligent transport systems decision-making system in real-time and to lower the requirements for peripheral equipment.