Ilya A. Sheremetiev1,
Alexander M. Kostetsky1,
Petr M. Yudanov1,
Igor S. Katorzhin1,
Ivan Yu. Sergeev2
1The Civil Defence Academy of the Ministry of the Russian Federation for Civil Defence, Emergencies and Elimination of Consequences of Natural Disasters, Khimki, Russia,
2Siberian Fire and Rescue Academy EMERCOM of Russia, Zheleznogorsk, Russia
This article is devoted to the applicability of neural networks for automatic analysis of images from CCTV cameras in real time for rapid response to emergencies (hereinafter referred to as emergencies) arising from landslides and mudflows. To design the architecture of the neural network and control its parameters, the high-level software platform tensorflow keras (hereinafter referred to as the platform) was used. The completeness metric (recall) at the level of 94.4% shows the success of the chosen approach. The results of this work can be used in the development of specialized software for the prevention of emergencies in rural areas and avalanche-prone regions and in the development of a digital system for monitoring the general situation of emergencies in peacetime, which will speed up the processing of information, as well as reduce the time for bringing operational information to the authorities and rescue units of the EMERCOM of Russia.
Open » // Ссылка для цитирования этой статьи