Abstract: Based on the basic concept of deep learning and the deep integration of surveying and mapping production technology, this paper explains the changes in the existing remote sensing interpretation production process brought about by the cross-border integration of surveying and mapping geographic information technology, and analyzes the challenges faced by intelligent production in the existing software and hardware environment. A new method of remote sensing interpretation production based on human-computer fusion intelligence is proposed, which breaks through multiple key technologies such as multi-GPU parallel training, rolling feedback training, and distributed micro-service applications, and developed a natural resource deep learning remote sensing intelligent solution. Translation platform and natural resources deep learning dynamic interpretation plug-in, and large-scale production applications have been carried out in key production links of surveying and mapping engineering projects such as the construction, maintenance and update of global geographic information resources. Through multiple projects, it has been verified that machine intelligence and human intelligence can be efficiently integrated through progressive human-computer interaction in remote sensing interpretation production, which greatly reduces the workload of production personnel and improves the scientificity and timeliness of remote sensing interpretation. It provides strong technical support for surveying and mapping production and natural resource survey and monitoring work.

Key words: remote sensing interpretation, human-machine integration, deep learning, surveying and mapping production, natural resources survey and monitoring