Detection of Group Test Objects Using Neural Networks with Hyperspectral Labeling During External Trajectory Measurements
Kuleshov I.A.
June 2022Springer
Measurement Techniques
2022#65Issue 3166 - 173 pp.
The task of detecting elements of group flight test objects is considered. The relevance of the research is related to the growing need for metrological support of external trajectory measurements when testing advanced multi-agent systems, which comprise aircrafts as elements. Existing algorithms for detecting objects used in modern optoelectronic stations (systems) are analyzed. According to the conducted analysis, the failure of these algorithms to meet the requirements imposed on testing group objects is largely due to the complexity of selecting objects against the atmospheric background. In order to ensure efficient detection of group objects, it is proposed to use additional highly-informative features that reflect the physical nature of test objects, such as the shape of the spectrum of light reflected from various materials. Such features can be identified in the process of preliminary hyperspectral analysis of test objects with a subsequent reduction in the data dimensionality. In order to eliminate errors arising during the rotation and partial overlapping of test objects, it is proposed to use neural network methods – in particular, the neural network detector YOLO v2. An approach to organizing data for training this detector is proposed along with an architecture that provides high accuracy and speed of information processing. These detector properties were confirmed experimentally on the basis of images exported from a model of external trajectory measurements. A procedure for implementing neural network detection of group test objects is proposed. The obtained results can be used when developing hardware and software for optoelectronic tracking systems.
convolutional network , detector , external trajectory measurements , group object , hyperspectral image
Text of the article Перейти на текст статьи
Ministry of Defense RF, Military Unit 03080, Priozersk, Kazakhstan
Ministry of Defense RF
10 лет помогаем публиковать статьи Международный издатель
Книга Публикация научной статьи Волощук 2026 Book Publication of a scientific article 2026