About the Possibility of Using Artificial Intelligence Technologies when Recognizing Images of Objects from Drones

Abstract

The problems associated with image recognition have long been solved within various practical tasks. This paper discusses the possibilities of using artificial intelligence technologies for recognizing images of objects from drones. The problem under consideration in this work is of such a type that it is necessary to recognize images of various objects belonging to the same classification and belonging to the same class; a certain number of objects in the image form a scene. In the developed system, a neural network will be used to identify and classify objects in real time from a video stream. The substantiation of the choice of the Caffe library for the system under consideration is given. Several options for constructing a neural network are considered. Empirically, it was found that for the neural network to work correctly, namely to work with the resulting video, the required height is 28.4 meters. After overcoming this height, the video quality deteriorates sharply, which greatly affects the classification of objects. The work of the classifier with other indicators is shown.