A Logical Approach to the Analysis of Aerospace Images

摘要

The paper proposes algorithms and software tools for the automatic interpretation and classification of objects and situations on aerospace images by structural-spatial analysis and iterative reasoning based on fuzzy logic and expert rules of inference. During iterations, the decision tree is built, the transition to local rules and additional features is carried out, and the ranges of acceptable values are adjusted. Particular attention is paid to geometric features of objects. Quantitative attributes are converted to qualitative ones for ease of perception of results and forming decision rules. The results of the experiment on the automatic identification of objects in the aerial image of an urban area are given. The system is useful for automating the process of labeling images for supervised learning and testing programs that recognize objects in aerospace images.