| Popis: |
[For the English abstract and full text of the article please see the attached PDF-File (English version follows Russian version)].The work was supported by the Russian Foundation for Basic Research (Grant No. 17-20-03034). ABSTRACT The article is devoted to development of a system for the intelligent analysis of video recordings of external surveillance cameras, which makes it possible to identify dangerous situations at railway facilities using the example of detection of falls in the track area. A method of preprocessing a video for the purpose of forming a feature space based on the use of background subtraction using the Gaussian mixture method, followed by tracking the movement of a person with the help of the Kalman filter and deformation of the shape of the mobile object as a result of applying the procrustean analysis is proposed. The selection of the optimal composition of the feature space and additional heuristics providing the isolation of episodes of falls from video recording with an average quality of the Cohen’s kappa 0,62 is compared with the visual analysis by the operator. Keywords: railway, safety, video surveillance, intelligent video analysis, motion recognition, machine learning, form analysis. |
