APPLICATION OF NEURAL NETWORKS TO DETERMINE THE MAIN CHARACTERISTICS OF ROAD TRAFFIC FLOWS IN THE CITY

Abstract

The article proposes a technique for automatic recognition and classification of vehicles based on the use of convolutional neural network Mask-R-CNN. The developed methodology makes it possible to automate obtaining information about the composition of the traffic flow and its intensity for each of the types of vehicles underlying any method of calculating emissions of pollutants by road. The article contains a description of the stages of neural network training, as well as the results obtained when using it. The proposed method of automating the process of assessing the intensity of the flow of vehicles, extracting data about it by analyzing the video stream from cameras installed on highways, and based on the use of the convolutional neural network Mask-R-CNN, showed good results. With the expansion of the test set, the number of classes of recognized cars can be increased, which will allow it to be used when applying methods for calculating emissions of pollutants that include a larger number of types of vehicles.