# AI satellite imagery analysis and GeoAI AI turns satellite images into information: a model learns from labelled examples what an object, a land cover type or a change looks like, then finds it across large areas automatically. Typical tasks are counting objects, classifying land cover and flagging changes. Models need examples that match your region and sensor, and their results need checking. ## How it works A satellite image is a grid of measured values. AI models learn from labelled examples which patterns in that grid stand for a building, a ship, a field or a clearing. Satellites now give a consistent, global view of forests, oceans and growing cities, but the images have little value without processing chains that turn pixel values into information for decision makers. Machine learning, especially deep learning, is increasingly part of these chains [1]. The models solve a few recurring tasks: ### Object detection and counting Because of the large volume of images collected every day, automatic detection and segmentation of objects are a constant need in Earth observation [2]. One example: the DeepSolar framework used deep learning on satellite imagery to locate solar installations and estimate their size across the contiguous United States [3]. Its use for solar power is described under [rooftop solar potential](https://leto.space/use-cases/rooftop-solar-potential/). ### Classification The model assigns each pixel or area to a class, such as forest, water, cropland or sealed ground. ### Change detection The model compares images of different dates and flags what has changed, for example new buildings or cleared forest. ### Foundation models These models are pre-trained on large data sets with self-supervised learning and then fine-tuned for a specific task. One example is Prithvi, released openly by NASA and IBM in 2023 and trained on Harmonized Landsat and Sentinel-2 imagery; it has been applied to flood mapping and to the detection of burn scars [4]. **Figure:** Three tasks on the same scene: detection finds and counts objects, classification assigns each area to a class, and change detection compares two dates. A foundation model is pre-trained on large data sets, fine-tuned with examples of your task and then tested on your region. Schematic scene. AI also runs on board satellites. ESA's Φsat-1, launched in 2020, was the first experiment to demonstrate AI for Earth observation in this way: it filtered out unusable images in orbit, so that only usable data was sent to Earth [5]. ## Which data and tasks | Task | Question it answers | Typical imagery | | ------------------------------ | ---------------------------------------------- | ---------------------------------------------------------------------- | | Object detection and counting | Where are the objects, and how many are there? | Very high resolution optical images, for example Pléiades Neo at 30 cm | | Classification | What covers the ground in each area? | Optical images such as Sentinel-2 at 10 m | | Change detection | What has changed between two dates? | Time series from the same sensor | | Fine-tuning a foundation model | Can a pre-trained model be adapted to my task? | Imagery similar to the model's training data | Which objects a model can find depends on the pixel size. Sentinel-2 records its finest bands at 10 m [6], enough for fields and forest stands but not for cars or small structures. Pléiades Neo records panchromatic images at 30 cm [7]. ## What you get - **Lists of detected objects** with their position and number, per site and date [2]. - **Classified maps**, for example of flooded areas or burn scars [4]. - **Change alerts** that point to the places where something new has appeared or disappeared. - **Information instead of pixel values**, produced by a processing chain that can run on each new image [1]. ## Limits **Limits** - **Training data must fit:** Objects look different from region to region and in images from different sensors, optical or radar, and platforms. Missing or scarce labels are a known limit of deep learning in Earth observation \[2]. A model trained elsewhere needs testing on your region and your imagery. - **Pixel size:** A model cannot find what the imagery does not resolve. Check the pixel size against the smallest object you need to see before you choose the imagery and the model. - **Results need checking:** Every model makes errors. Before a result feeds a decision, compare it with reference data you trust, and keep checking samples once the model is in use. ## How we help Whether a ready service, a platform or a custom model fits depends on your question. In a [strategy engagement](https://leto.space/consulting/strategy/) we work out which question the model must answer, how accurate it must be and which imagery it needs, and we then write the requirements for choosing a provider. Spacegraph, our discovery platform, has marketplace categories for object detection and counting and for change detection. For teams that want to build skills in-house, the LETO Academy offers the course ["GeoAI: Deep Learning for Earth Observation"](https://academy.leto.space/en/courses/geoai-earth-observation), at intermediate level and by invitation, with exercises in Python. [LETO Academy](https://academy.leto.space): Our self-paced courses on AI, digital skills and space-enabled solutions, in English and German, some open, some by invitation. [Skills and workshops](https://leto.space/consulting/workshops/) ## Related use cases - [Analysing global outdoor storage sites](https://leto.space/use-cases/outdoor-storage-analysis/): Activity at storage sites as an input to supply-chain planning. - [Tracking land use change](https://leto.space/use-cases/land-use-change/): Change detection applied to land take and sealing. - [Monitoring forest health and deforestation](https://leto.space/use-cases/forest-monitoring/): Alerts for clearings and stressed forest stands. ## Frequently asked questions ### What is GeoAI? GeoAI is the use of machine learning on geographic data such as satellite and aerial images. Satellite data has little value without processing chains that turn pixel values into information for decision makers, and machine learning, especially deep learning, is increasingly part of these chains [1]. ### Do we need to train our own model? Not always. Foundation models such as Prithvi, released openly by NASA and IBM in 2023, are pre-trained on large data sets and can then be fine-tuned for a specific task [4]. Fine-tuning still needs examples from your task, and the result needs a test on your region. ### How do we start a pilot with AI on satellite imagery? Define one question the model must answer, for example how many objects of a type stand on a set of sites, and collect a set of examples you trust to check the results. Then compare an existing service with a model adapted to your data. ## Sources 1. [Artificial Intelligence to Advance Earth Observation: A review of models, recent trends, and pathways forward (Tuia, Schindler, Demir, Zhu et al.), IEEE Geoscience and Remote Sensing Magazine, 2024-09-09](https://doi.org/10.1109/MGRS.2024.3425961) 2. [Remote Sensing Object Detection in the Deep Learning Era: A Review (Gui, Song, Qin and Tang, Remote Sensing 16, 327), Remote Sensing (MDPI), 2024-01-12](https://doi.org/10.3390/rs16020327) 3. [DeepSolar: A Machine Learning Framework to Efficiently Construct a Solar Deployment Database in the United States (Yu et al., Joule 2, 2605 to 2617), Joule, 2018-12-19](https://doi.org/10.1016/j.joule.2018.11.021) 4. [IBM Research and NASA IMPACT release open-source geospatial foundation model on Hugging Face, NASA Earthdata, 2023-08-03](https://www.earthdata.nasa.gov/news/blog/ibm-research-nasa-impact-release-open-source-geospatial-foundation-model-hugging-face) 5. [Φsat (mission page), European Space Agency (ESA), accessed 2026-10-08](https://www.esa.int/Applications/Observing_the_Earth/FutureEO/Phsat) 6. [Sentinel-2 mission, European Space Agency (ESA), SentiWiki, accessed 2026-10-06](https://sentiwiki.copernicus.eu/web/s2-mission) 7. [Pléiades Neo, European Space Agency (ESA), Earth Online, accessed 2026-10-06](https://earth.esa.int/eogateway/missions/pleiades-neo) Page: