AI+ Remote Sensing Technology Empowers 3D Real Scene China

Relying on decades of software development experience in the field of geographic information, SuperMap continues to carry out research and development innovations in technologies related to the construction of 3D real scene China, and provides basic product support for the implementation of the task of building a 3D real scene China through AI+remote sensing technology.

3D Real Scene

3D Real Scene is an important new national infrastructure that reflects the spatio-temporal information of human production, life and ecological space in a real, three-dimensional and time-sequential manner. It realizes real-time connection and interconnection between digital space and real space through "human-machine compatibility, Internet of Things perception and ubiquitous services", and provides a unified spatial positioning framework and analysis basis for digital China. It is an important strategic data resources and production factor for digital government and digital economy.

The construction of terrain-level 3D real scene includes: constructing terrain-level geographic scenes and basic geographic entities, obtaining other entities and IoT perception data, assembling and generating terrain-level 3D real scene products for 3D visualization and spatial measurement, and serving the macro-planning. Among them, terrain-level geographic scenes include remote sensing image products such as digital elevation models (DEM), digital surface models (DSM), digital orthophotos (DOM), and true orthophotos (TDOM) .

Introduction to AI GIS Technology

In 2019, SuperMap released SuperMap GIS 10i, a product based on the artificial intelligence GIS (AI GIS) technology system. After multiple versions, AI GIS technology and products are becoming more mature. Diversified products can meet the personalized needs of different users. Its core content includes:

GeoAI: Spatial analysis and processing that integrates AI. Based on basic theories and algorithms of artificial intelligence such as machine learning and deep learning, the software implements a series of GeoAI functions for fields such as natural resource monitoring and urban planning, and provides remote sensing image intelligent interpretation tools such as target detection, binary classification, land feature classification, scene classification, and object extraction based on deep learning.

AI for GIS: AI empowers GIS, that is, based on AI technology, GIS software functions are enhanced and optimized. By applying AI technology to some traditional GIS functions, the intelligent evolution of GIS software functions can be achieved. For example, the AI attribute collection function can efficiently identify business attribute data such as illegal parking, small advertisements, and manhole covers based on AI technology; the AI mapping function can provide lower-cost and more convenient indoor mapping services; AI+AR technology can integrate AR technology into the AI system to achieve a visualization effect that combines both virtual and real.

GIS for AI: GIS empowers AI, that is, based on GIS technology, AI analysis results are further processed, analyzed and spatially visualized. For example, in map visualization applications such as traffic flow monitoring, urban parts and case management, it can provide decision-makers with a more intuitive form of information expression; in addition, it can also monitor pedestrians and motor vehicles entering the bus lane based on video monitoring data through geo-fencing, video spatial computing and other technologies.

AI+ remote sensing technology empowers the production of basic geographical entities

SuperMap's artificial intelligence GIS technology system also includes the full-process technology of intelligent interpretation of remote sensing images based on deep learning, and provides remote sensing data preprocessing tools, sample management tools, training sample production tools, model training tools, intelligent interpretation tools, post-processing tools, etc., which can effectively support the production of basic geographical entities.

Among them, intelligent interpretation tools include:

• Target detection tool: This tool uses deep learning algorithms to automatically determine and identify the category and location of one or more targets in remote sensing images, and uses vector boxes as markers to help users identify different ground entities such as ports, bridges, and electric towers.
• Scene classification tool: This tool can distinguish images with similar scene features from multiple images and assign a scene type label to each image.
• Binary classification tool: This tool is based on a deep learning network model. By analyzing the spectral information of various types of objects in remote sensing images, it can extract single land types with obvious characteristics such as buildings, roads, and rivers.
• Object classification tool: This tool classifies and identifies the feature information of remote sensing images and divides them into a series of image regions with specific semantic information.
• Object extraction tool: Based on classification, this tool can extract independent vector objects from the same category of features.