Search

With the rapid development of geospatial science, artificial intelligence, and high-performance computing, geospatial artificial intelligence has become the major technique for processing and analyzing geospatial big data, and will be widely used in the scientific research and engineering application of earth science, spatial cognition, and smart city. Geospatial intelligence, as an inter-discipline between geospatial science and artificial intelligence, is driven by dual disciplines. At present, it has made important progress in hardware research, system development, data and model sharing, services and applications, and is also facing new challenges. This paper first describes the conceptual evolution of geospatial artificial intelligence and lists the frameworks of some technical systems. The paper then reviews the state-of-the-art in the scientific research and typical application domains. The problems and challenges facing geospatial intelligence are analyzed. Finally, some future prospects and trends are presented.

Carrying out an integrated survey, monitoring and evaluation of natural resources, accurately understanding the status and changes of various natural resources in China, is the scientific basis for territorial and spatial plans, and gradually realizing the overall protection, restoration, and comprehensive management of landscapes (including mountains, forests, fields, lakes and grasses), ensuring national ecological security. At present, the extraction of natural resource features based on remote sensing images mainly relies on visual interpretation via man-machine interaction and field spot verification. It needs high labor intensity, and production efficiency is low. The results are also highly affected by man-induced factor, which can no longer adapt to the requirements for integrated investigation and monitoring of all features of natural resources. This paper conforms to the emerging direction of the research development with artificial intelligence collaboration. Firstly, this paper reviews the main research progress of deep learning technology and its application in the field of remote sensing image intelligent extraction systematically, and analyzes its limitations, then it reviews the main research status of man-machine collaboration. Afterward, Starting from the characteristics of natural resource features, presents a technical framework of "intelligent background computing+intelligent engine+man-machine interface" for man-machine collaborative intelligent extraction. The key technologies that need to be overcome are described. At last, the idea of creating cloud platform for feature extraction are discussed. This paper aims to provide a new AI method for intelligent extraction and improve the automation and intelligence level of natural resource feature extraction.

The history and recent development of photogrammetry and remote sensing are reviewed and analyzed firstly. Then the novel concept of "photogrammetric remote sensing" is put forward to meet the needs of accurate and fast processing of multi-source remote sensing images in the new era of big data and intelligent surveying and mapping. The new photogrammetric remote sensing discipline is the deep integration of the frontier theories and technologies of photogrammetry and remote sensing, and concentrates on solving the theories and technologies about simultaneously determine the geometric positions, physical attributes, semantic information and temporal changes of interested scenes and objects. Its theoretical and fundamental basis are photogrammetry, remote sensing, artificial intelligence, big data processing, and high-performance computation, etc. It will break through the current isolated status and serial technical route that photogrammetry mainly focuses on geometric processing, while remote sensing mainly focuses on semantic information extraction and inversion. It forms an innovation of the closed-loop fusion of semantic extraction and geometric processing. A novel geometric-semantic integrated processing framework is formed through the deep fusion of geometric model and spectral radiative and reflective information. Based on the proposed concept of photogrammetric remote sensing, this paper discusses the main scientific problems and related research and application fields, and then attempts to build a new theoretical and technical framework of integrated intelligent photogrammetric remote sensing processing of multi-source satellite images. Closed-loop fusion of semantic information extraction and accurate geometric processing has significantly improved the level of accuracy, automation and intelligence. The correctness and effectiveness of the proposed theory and methods are preliminarily verified by several practical applications.

Reliability is one of the important features in remotely sensed data-based land use monitoring. Artificial intelligence (AI) technology promotes the rapid development of object recognition from remotely sensed data. However, the un-explainability in such image processing causes reliability problems. Based on the reliability theory and the basic theory of AI, this paper first presents the idea and the overall framework of intelligent and reliable object recognition. Second, the core research directions, including analysis of influencing factors, improvement methods, evaluation methods, and process control for reliability are sequentially introduced. Finally, the future development trend of AI for reliable object recognition from remotely sensed data is outlined.

In recent years, point cloud has become an important type of 3D spatial data. How to improve the understanding abilities of point cloud using artificial intelligence for correct semantic labeling and accurate detection of objects is an urgent and difficult problem. This paper hence proposes an end-to-end 3D point cloud deep learning network, which effectively guarantees the efficiencies of point cloud sampling, the accuracy of feature extraction and the optimization of the overall network performance by the up-down sampling strategy of irregular distribution point cloud, multi-layer aggregation and propagation of features and the loss function for uneven samples. The studies on the large-scale 3D point cloud benchmark data show that it achieves excellent performance in semantic labeling for large-scale outdoor scenes of point clouds, better than those of the state-of-art deep learning networks of point cloud, providing a strong support for the high-performance extraction of 3D geospatial information.