Remote sensing is a core, cutting-edge technology that analyzes satellite, aerial, and drone imagery to achieve land cover, environmental monitoring, geospatial understanding, and Earth observation. It primarily addresses the challenges of "lack of macroscopic perspective and difficulty in data acquisition" caused by geographical limitations in human Earth observation. This architecture analyzes satellite, aerial, and drone imagery to acquire land cover and environmental information without physical contact with the target, achieving a simultaneous closed-loop geospatial understanding without the need for on-site human reconnaissance. It typically utilizes optical, radar, infrared, and hyperspectral sensors mounted on satellites, aircraft, and drones to receive electromagnetic waves reflected or emitted by targets and analyze the data to identify land features and monitor environmental changes.

Remote sensing was not originally a concept proposed in the field of artificial intelligence, but rather originated from earth sciences, surveying, and spatial information. In the 1950s, geographers at the US Office of Naval Research (ONR)... Evelyn L. Pruitt The term "Remote Sensing" was coined to describe methods for acquiring information about the Earth without physical contact. Evelyn L. Pruitt's concept of Remote Sensing is not directly related to today's "remote sensing" in the field of AI. At that time, artificial intelligence and deep learning had not yet developed, and remote sensing mainly relied on aerial photography, physical models, manual interpretation, and traditional statistical methods for analysis. With NASA's launch of Earth observation satellites such as Landsat-1 in 1972, remote sensing gradually developed into the modern Earth Observation system, becoming a crucial technology for acquiring global environmental, resource, and geographic information.

Since the breakthrough of Convolutional Neural Networks (CNNs) in computer vision in 2012, AI has been widely applied to tasks such as remote sensing image classification, object detection, semantic segmentation, and change detection. In recent years, with the rise of large-scale model technology, research has further developed into areas such as Remote Sensing Foundation Models, GeoAI (Geospace Artificial Intelligence), and multimodal Earth observation models.

Representative research achievements in the field of artificial intelligence remote sensing include: AlexNet in 2012, which promoted the development of deep learning visual models and provided a technical foundation for intelligent analysis of remote sensing images; Vision Transformer (ViT) in 2020, which introduced Transformers into visual tasks and promoted their application in remote sensing scenarios; and the joint launch by IBM and NASA in 2023. Prithvi Earth observation fundamental models, which apply large-scale pre-trained models to satellite remote sensing data analysis, have become an important representative of remote sensing fundamental models. Related papers include... Foundation Models for Generalist Geospatial Artificial Intelligence(2023)wait.