Leveled Approach of Context Setting in Semantic Understanding of Remote Sensing Images

Geographic data is carried via high spatial resolution (HSR) remote sensing (RS) images. Understanding HSR remote sensing images requires a thorough exploration of geo-objects and their spatial relationships. The task of semantic understanding is divided into four subtasks preprocessing, object detection, identifying geospatial relations, and semantic description generation. These semantic descriptions will be of great significance in the field of urban planning, RS image retrieval, military intelligence, etc. A leveled approach for context setting is proposed in this paper for the semantic understanding of RS images. This approach comprises various steps such as scene classification, multi-label classification, object detection and labeling, description generation using geospatial relationships, and finally generating detailed descriptions for input RS images. Results show that the proposed method performs well in generating a description that considers the overall context of the RS scene.