Method of automating the process of object detection to increase the efficiency of deciphering aerial reconnaissance data

Abstract

Problem factors that have a significant impact on the effectiveness of the process of deciphering air reconnaissance data obtained under the conditions of the use of unmanned aerial vehicles of the tactical level are analyzed, taking into account the experience of conducting combat operations on the territory of Ukraine. Requirements for operational efficiency and reliability of the process of recognition (detection) of aerial reconnaissance objects are being formed. Modern computer vision algorithms and deep machine learning technologies are studied from the point of view of integration into the process of deciphering aerial reconnaissance data. A method of automating the process of detecting aerial reconnaissance objects is being developed to increase the effectiveness of deciphering aerial photographs. The essence of the developed method is the integration of modern computer vision technologies, implemented on the basis of algorithms of artificial neural networks, in the process of deciphering aerial photographs. Analysis of the effectiveness of the developed method from the point of view of ensuring the required level of reliability shows that the use of the developed method allows obtaining the required values of the probability of recognition of aerial reconnaissance objects on aerial photographs. In turn, the analysis of quantitative evaluations of the efficiency of deciphering aerial reconnaissance data shows that the use of the developed method allows you to significantly reduce the time for deciphering an aerial photograph on average, up to 2 seconds per aerial photograph under the conditions of using rather low computing power. The features of the developed method: practical implementation of the specified automation method does not require high computing power; integration into the decryption process does not require significant personnel training and high technical requirements.