In the era of ubiquitous connectivity, remote sensing satellites serve as a critical means of spatial data acquisition. Converting massive volumes of remote sensing imagery into high-quality standardized base data and rapidly extracting actionable insights to support scientific decision-making has long been a pursuit for both academia and industry.
As spatial data grows exponentially, industries urgently demand efficient, intelligent solutions—with artificial intelligence (AI) emerging as the key driver of innovation.
Over the past two years, SuperMap has released cross-platform desktop and server software for remote sensing image processing, enabling end-to-end unified remote sensing and GIS workflows. This advances rapid application of remote sensing data products within one day (T+1).
Currently, SuperMap is deeply integrating AI to develop next-generation intelligent, high-precision remote sensing software with unified GIS capabilities. This integration streamlines the entire workflow—from data acquisition, processing, and analysis to management and application.
01. AI + Intelligent Remote Sensing Image Production
High-quality AI outputs demand high-quality spatial data inputs. In the AI era, automated feature change detection requires sub-pixel feature alignment across historical imagery—a challenge beyond traditional processing techniques. How does next-gen AI-powered remote sensing software address this?
02. Out-of-the-box Pretrained Models for AI Interpretation
To enhance accessibility, SuperMap offers out-of-the-box pretrained models for:
- Urban water extraction
- Building extraction
- Farmland identification
- Greenhouse detection
- Forest cover mapping
- Cloud exaction
Faced with rapidly growing demands, users seek expanded feature recognition capabilities. Are there any new advancements?
03. Large-Scale Remote Sensing Interpretation Models
While CV-domain foundation models (e.g., SAM) inspire new technical directions, most are trained on everyday images and underperform on remote sensing tasks.
Addressing this, SuperMap launched its proprietary LIM model in June 2024. LIM’s architecture combines:
- An upstream self-supervised learning network
- A downstream classification task network
enabling single-pass extraction of eight feature types.
Amid rapid advances in self-supervised training and abundant unlabeled imagery, how does LIM achieve breakthroughs? How can such large models efficiently adapt to diverse downstream tasks?
On June 24-25, all the answers will be revealed at the 2025 Geospatial Intelligence Software Technology
Conference in China National Convention Center in Beijing!
Theconference will focus on cutting-edge technologies and industry hotspots, showcasing the latest advancements in geospatial intelligence software technology and products, and exploring new opportunities for industry development. We look forward to your presence!
