Search and recognition of satellites in Earth orbits

The study is dedicated to the problem of detecting and identifying spacecraft in the image for analyzing their state by visual inspection. The paper presents solution for the search and recognition of satellites in images of outer space using neural networks. To implement the software package, open sources are used: processed and prepared photographs of satellites in near-Earth orbits, a marking tool — naming an object in a photograph, as well as the Detectron software system that implements modern object detection algorithms. The paper presents a description of the collected architecture and the prepared data set for training and testing the neural network. The results of testing different architectures are presented and possible options for its refinement to solve the intare problem.