ИССЛЕДОВАНИЕ ВЛИЯНИЯ СЛОЖНЫХ УСЛОВИЙ ПОЖАРНОЙ ОБСТАНОВКИ НА КАЧЕСТВО НАБЛЮДЕНИЯ И БЕЗОПАСНОСТЬ ПОЛЁТА БЛА

Abstract

Aviation monitoring of fires with the help of unmanned aerial vehicles (UAVs), in particular,forest ones, during which the search for various objects of interest is carried out: people, cars,etc., is one of the most effective measures to reduce the level of possible losses. In this paper, weconsider approaches to the formation of algorithms for processing and improving images obtainedin the process of monitoring the fire situation, based on the use of neural networks, as well asimage filtering algorithms, in order to search for various objects of interest. Fire monitoring usingan UAV is a two-criteria task: there is a need to protect the device from the thermal effects of thefire as much as possible, as well as to maximize the observability, which can be achieved by reducingthe altitude of the flight. This paper presents the empirical models developed by the authors forthe flight safety of an unmanned aerial vehicle and the observability of objects of interest in theprocess of monitoring the fire situation. The proposed models allow us to take into account thefeatures of the monitoring conditions, such as the priority of detecting the object of interest to thesecurity of the reconnaissance vehicle itself, air humidity, terrain and type of terrain, time of day,and so on. An example of the application of the contrast model is considered on the example of thesearch and detection of the "letter"label. On the basis of the conducted experiment on the recognitionof the mark in the smoke, the analysis of the proposed models is carried out, the quantitativeresults are given. The paper describes the criteria for the optimal choice of the altitude of theflight of the device over the observed scene, which are formed on the basis of the base of expertassessments, as well as the proposed models of the observability and safety of the UAV flight. Dependingon the target search task, the optimality criterion for choosing the UAV flight altitudeover the observed scene may vary.