Industrial Automatic Control Systems and Controllers Annotation << Back Training and Use of a Convolutional Neural Network for the Detection and Classification of Vehicles in Ultra-high Resolution Satellite Imagery I.N. Pugachev, G.Ya. Markelov, V.S. Tormozov The paper focuses on the application of a convolutional neural network to detect and classify vehicles on ultra-high resolution satellite images. The purpose of the work is to develop SNS and to establish parameters of its architecture for detection and classification of vehicles on satellite images of ultra-high resolution. The elements of novelty of the presented solution are the use of CNN for the task of recognition of vehicles on ultra-high resolution satellite images, identification of optimal parameters of CNN for the solved task. The practical significance is that the CNN of the proposed architecture and with the identified parameters, through the use of the city ‘s traffic fl ow assessment system, will improve the factual support of transport planning processes and improve the quality of traffic management systems. Keywords: transport planning; satellite imagery; ultra high resolution; convolutional neural network; image processing; digital image processing; machine learning; satellite imagery; ultra high resolution; image recognition. DOI: 10.25791/asu.10.2019.933 Contacts: - Pp. 20-25.