Application of Dynamic Graph CNN* and FICP for Detection and Research Archaeology Sites

The paper proposes a methodology for solving the task of accurate semantic classification of 3D data using a combination of 2D and 3D methods based on the YOLO detector and the modified DGCNN network. The methodology is tested on the example of the problem of classification of large-scale geospatial objects, such as digital relief models of archaeological sites. A method for accurate registration of objects (FCIP) in the class of affine transformations using geometric and color features was proposed. The results of computer modeling of the proposed methodology based on FICP+DGCNN*+YOLO were presented and discussed. The methodology has theoretical and applied significance not only for the decryption and research of archaeological sites, but also for many applications of digital information processing and robotics in general.