On the morning of May 22nd, the 2025 Global Spatio Temporal Intelligence Conference (WGDC25) Satellite Application Innovation Forum was held in Beijing.
At the meeting site, experts and scholars including Meng Yu, director of the National Research Center for Remote Sensing Application Engineering Technology, Gu Haiyan, deputy director of the Natural Resources Survey and Monitoring Research Center of the Chinese Academy of Surveying and Mapping Sciences, Li Suju, director of the Satellite Application Department of the National Disaster Reduction Center of the Ministry of Emergency Management, Cheng Bo, researcher of the Aerospace Information Innovation Institute of the Chinese Academy of Sciences, attended the round table forum to carry out in-depth discussions around "from single to multimodal: the evolution of satellite remote sensing AI", and propose a new path for the breakthrough and application of remote sensing technology.
Gu Haiyan pointed out that multimodal fusion needs to overcome technical difficulties such as data alignment and higher performance for complex data, and called for the application of AI interpretable research techniques in the remote sensing industry. At the same time, AI is not omnipotent and cannot solve all problems when facing business scenarios. "We need to deeply understand and research the needs of various places, constantly experiment from the perspectives of content indicators, technical routes, etc. After successful experiments, we can promote and apply them
Taking landslide rescue in Guizhou as an example, Li Suju pointed out that there are more uncertainties in emergency scenarios, and data acquisition, processing, and analysis need to be carried out in a short period of time. Due to the insufficient sample size of low probability events and the complexity of the scenarios themselves, the generalization of current key technology research and development methods is limited. She emphasized that she hopes to use scenario driven approaches in applications for multi-party iteration.
Cheng Bo stated that China's remote sensing satellites have developed rapidly, and the data model has evolved from early single type to current multimodal. At the same time, information diversity poses challenges to data analysis, and it is hoped that intelligent means such as AI and AGI can provide more technical support for multimodal fusion. At the same time, he called for data sharing, product sharing, and technology sharing, "so as to promote the rapid development of AI intelligence and meet the AI era faster. ”
