News Release

A spatiotemporal intelligent framework and experimental platform for urban digital twins

Peer-Reviewed Publication

Beijing Zhongke Journal Publising Co. Ltd.

The GeoAI Framework for urban Digital Twins

image: The GeoAI framework includes entity object management, virtual space construction, IoT perception, intelligent computing analysis, intelligent central decision-making, and application services. view more 

Credit: Beijing Zhongke Journal Publising Co. Ltd.

Research Background

The era of Big Data features intelligence, ubiquity, and interconnection of all things. It comes with other advanced information technologies, such as the Internet, Cloud Computing (CC), Internet of Things (IoT), and Artificial Intelligence (AI) . Human society has also gradually entered the Ternary Space from the Binary Space. That is, from the Social Space (the sum of human behavior and social activities) to the Information Space (the computer, Internet, and data information built on physical space and social space) . The Ternary Space maps and digitally connect the urban physical and social space to the information space, thus promoting urban intelligence. This technological evolution empowers the Digital Twins Technology (DTT) and expands its application from the industrial field to Smart City construction, namely the Digital Twins of Smart Cities. DTT originates from Virtual Simulation, and along with parallel system technology, it can replicate and reconstruct real-world objects, systems, and processes in the virtual information space. The digital and physical visions collectively form the Digital Twins. Digital Twins of Smart City constructs an urban twin model in the virtual space, which coexists and blends with its physical counterpart.

Unarguably, China is deemed by many the world's safest country, and the Chinese government emphasizes its national and citizen's security like no other country. This can be well reflected in cameras installed in public areas, constituting a video "Skynet" covering key areas and industries. Video Big Data can visualize the human world. In Smart Cities, it can map, analyze, and predict the urban state through fusion technology and multi-dimensional data association analysis in the urban Digital Twins. AI technology must come to play to mine the video Big Data. Spatiotemporal information is the infrastructure of urban Digital Twins. The Geographic Information System (GIS) maps physical entities and social relations to information space by digital projection. It lays the foundation of urban ternary spatial association and Smart City construction. Meanwhile, Geospatial Artificial Intelligence (GeoAI) can enhance the dynamic perception, intelligent reasoning, and knowledge discovery capabilities of urban geographic phenomena and earth science processes. Thereby, it boosts the intelligence of geospatial perception, understanding, cognition, and decision-making. Spatiotemporal intelligence provides a space-time base for constructing urban Digital Twins and improves the geospatial correlation, Virtual-Real Fusion (VRF), Augmented Reality (AR) interaction, and intelligent decision-making collaboration in the urban Ternary Space. However, the accuracy of the existing spatiotemporal intelligent algorithms about Digital Twins is not high. With the increase of the data volume, the operation of the model spatiotemporal cube becomes more complex. In addition, the data redundancy problem of existing algorithms has not been effectively solved. Therefore, it is necessary to find a suitable intelligent spatiotemporal architecture scheme to optimize the target detection.

This work reviews the relevant literature, conducts experimental verification, studies the development status of distributed computing centers, and targets the existing problems. The geographical framework of distributed computing center is creatively proposed. The main contribution is to propose an FSSiamese target tracking algorithm based on Digital Twins network. Then, a GeoAI experimental platform is built to verify the key technologies of accurate spatial-temporal video analysis. The finding can provide new research ideas for constructing distributed control systems.


Results and Significance

Urban Digital Twins are a way of urban intelligence under the Ternary Space theory. It maps urban physical space and social space to information space to discover and analyze problems and find solutions. Then, it furnishes physical and social space with feedback to solve urban problems. Geospatial intelligence abstracts physical and social space from the geospatial perspective and improves the intelligence level of urban geospatial perception, understanding, cognition, and decision-making. It establishes the relationship of 3D space. This work promotes the innovation and optimization of geospatial intelligence technology by constructing a GeoAI framework and experimental platform for urban Digital Twins through intelligent analysis of urban VRF, decision-making, and practical feedback. Then, technical solutions are provided for constructing urban Digital Twins.

This paper validated the effectiveness of the proposed improved small object detection model and twin network video tracking model through experiments. The proposed YOLOv5-Pyramid small object detection model showed more precise performance with an average precision value higher than other two methods, while the FSSiamese object tracking algorithm based on the twin network achieved higher accuracy and faster inference speed. In addition, the study proposed a multi-scale feature extraction module and a local population density map output method with geographical reference, which enabled the statistical analysis of population in a region. Furthermore, the study used various geometric transformation methods for the georeferencing correction, and demonstrated different results through experiments. The paper pointed out that more experiments are needed to verify the accuracy of the study, which is also the future work for further research.


Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.