Российский технологический журнал (Apr 2022)
Semantics of visual models in space research
Abstract
Objectives. The aim of the study is to develop a methodology for assessing the semantics of weakly structured or morphologically complex visual information models. In order to achieve the goal, a criterion for classifying visual models as complex and an algorithm for obtaining a gradient image with several levels of density were introduced. The gradient image is not binary, thus increasing the reliability of finding boundaries or contours. An auxiliary structural visual model was introduced, and a series of images of different densities was used in processing. Next, the concept of a conditional image coordinate system was introduced. This allows for information to be transferred from different visual models to a synthetic resulting visual model.Methods. Using gradient image processing and constructing a new intermediate structural model allows models with different densities to be linked. A system of conditional image coordinates was introduced and a series of models with different densities to obtain a synthetic image was processed.Results. The visual models obtained from satellite images with poor visibility of objects were processed in the Sun– Earth–Moon system. The Sun–Earth system was chosen as the basis. A characteristic of space images is the fact that the bright light of the Sun “clogs” the images of other objects with large phase angles. The use of the contouring technique allows for the visibility of images of low brightness and high brightness to be equalised. The shift of the frequency response after detection of all objects enabled the formation of a clear visual model.Conclusions. In primary visual models, low brightness images were not visible. They appeared when exposure was increased, while high-density objects merged into one. Because of this, it is fundamentally impossible to obtain a high-quality image of all objects, or the complete semantics of a visual model from a single high, medium, or lowdensity image. In order to obtain the complete semantics of the visual model, a series of images need to be processed with the transfer of images to a common synthetic image. The proposed technique allowed for such problems to be resolved. A comparison of the results obtained using the methods of processing a single image proved the reliability and high information content of the method.
Keywords
