REMOTE MONITORING OF MINES IN FIELDS WITH USING NEURAL NETWORKS

Abstract

The work is devoted to the issue of humanitarian demining of agricultural lands. The difference between humanitarian demining and military demining is the lack of strict deadlines for the implementation of measures, as well as relatively small human and technical resources and large areas of monitoring. The situation is particularly difficult on marginal lands, which are poorly monitored by the local population and, accordingly, the establishment of minefields may be unexpected for farmers. Marginal lands are a potential source of energy raw materials for bioenergy, the importance of which is increasing in the modern conditions of the destruction of traditional logistics chains of supply of energy resources. The purpose of the work was to develop a method of express monitoring regarding the presence of mines in the fields. Operational research of large areas is possible precisely thanks to remote monitoring technologies, in particular thermal imaging. The hypothesis about the possibility of identifying mines due to the difference in the dynamics of heating and cooling of explosive objects under solar radiation during the day was tested. The research was conducted on the basis of the training ground of the military department of the National Academy of Sciences of Ukraine. TM-62M anti-tank mines, high-explosive shells caliber 125 mm, etc. were studied. When setting up the experiment, part of the ammunition was buried in the soil to a depth of 2-5 cm in accordance with the instructions for demining. It was established that only sunny days are suitable for thermal imaging monitoring, and samples installed in the shade were not identified. The maximum temperature difference for mines was recorded in the morning, and for massive shells in the evening. With regard to buried mines and projectiles, during thermal imaging monitoring, only depressions and bumps on the site were recorded. Based on the low selectivity of thermal imaging monitoring, neural networks were used to indicate mechanically laid mines on the ground. Positive results were obtained.