Applications of Artificial Intelligence in Remote Sensing

Forest carbon monitoring gets an AI boost, reports Abhishyant Kidangoor. Forests have long been surveyed from above. Satellite data reveal where they stand and how they shrink or grow, while lidar—laser-based radar—has allowed scientists to map them in 3D, uncovering details that lie beyond human sight. Now, artificial intelligence is adding a new layer of insight. Earth-imaging company Planet has unveiled a Forest Carbon Monitoring tool that fuses its satellite imagery with lidar data. The tool can estimate carbon storage, tree height, and canopy cover in remote forests at a granular resolution of three meters. “It will help us understand aspects of the forest that might not be initially accessible to the naked eye,” says Andrew Zolli, Planet’s chief impact officer. Satellites track forest cover but not the carbon stored in biomass. Measuring this requires lidar, which calculates tree dimensions by measuring the time laser beams take to bounce off foliage. NASA’s GEDI mission, mounted on the International Space Station, has mapped swathes of forests, but coverage gaps persist. Planet’s tool aims to bridge these voids, training machine-learning models to infer carbon data in areas without lidar coverage. Initial findings from the tool have been striking. While deforestation ravages the Amazon, the northern reaches harbor untouched carbon reserves. “What really resonated with me is the understanding of where we have extant forest carbon stocks which we must absolutely protect,” says Zolli. The data also underpin Project Centinela, which supports conservation efforts in biodiversity hotspots like Tanzania’s Gombe Stream National Park. Meanwhile, carbon markets—often criticized for opacity—may gain credibility through applications of the tool argues Zolli: “The data gives a shared, common picture of what’s actually happening on the ground.” Planet’s innovation rests on decades of data, cutting-edge AI, and cloud computing. “We are the first generation that has had all three in place,” Zolli says, enabling swift, confident assessments of carbon across the globe. 📰 story: https://lnkd.in/gwRWf5Qf 📷: A view of carbon storage in forest and an area of fishbone deforestation in the Brazilian Amazon. Image courtesy of Planet.

𝐇𝐲𝐩𝐞𝐫𝐅𝐫𝐞𝐞: 𝐀 𝐅𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧 𝐌𝐨𝐝𝐞𝐥 𝐟𝐨𝐫 𝐇𝐲𝐩𝐞𝐫𝐬𝐩𝐞𝐜𝐭𝐫𝐚𝐥 𝐈𝐦𝐚𝐠𝐞𝐫𝐲 What if an AI could analyze any hyperspectral image - from 46 to 274 spectral channels - without requiring image-by-image tuning? Even better, what if it could identify multiple objects of the same class with just a single prompt? Jingtao Li et al. introduced HyperFree at CVPR 2025 - the first channel-adaptive, tuning-free foundation model for hyperspectral remote sensing that achieves comparable results to specialized models using just one prompt versus their five training shots. 𝐓𝐡𝐞 𝐡𝐲𝐩𝐞𝐫𝐬𝐩𝐞𝐜𝐭𝐫𝐚𝐥 𝐜𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐞: Hyperspectral imagery captures hundreds of spectral channels spanning 400-2500nm, providing incredibly detailed Earth observation data. However, existing foundation models face a major bottleneck - they require expensive fine-tuning for each new dataset due to varying channel numbers across different sensors. This makes deployment costly and time-consuming, especially problematic given the high acquisition costs of hyperspectral data. 𝐊𝐞𝐲 𝐢𝐧𝐧𝐨𝐯𝐚𝐭𝐢𝐨𝐧𝐬: ∙ 𝐋𝐞𝐚𝐫𝐧𝐚𝐛𝐥𝐞 𝐰𝐚𝐯𝐞𝐥𝐞𝐧𝐠𝐭𝐡 𝐝𝐢𝐜𝐭𝐢𝐨𝐧𝐚𝐫𝐲: Dynamic embedding generation spanning full spectrum (400-2500nm) at 10nm intervals, allowing processing of any hyperspectral sensor ∙ 𝐏𝐫𝐨𝐦𝐩𝐭-𝐦𝐚𝐬𝐤-𝐟𝐞𝐚𝐭𝐮𝐫𝐞 𝐢𝐧𝐭𝐞𝐫𝐚𝐜𝐭𝐢𝐨𝐧: Novel approach treating feature distance as semantic similarity to generate multiple masks from a single prompt ∙ 𝐇𝐲𝐩𝐞𝐫-𝐒𝐞𝐠 𝐝𝐚𝐭𝐚 𝐞𝐧𝐠𝐢𝐧𝐞: Automated generation of 50k high-resolution hyperspectral images with 15 million segmentation masks using Segment Anything ∙ 𝐓𝐮𝐧𝐢𝐧𝐠-𝐟𝐫𝐞𝐞 𝐚𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞: Direct processing of unseen datasets across five tasks without model modification 𝐖𝐡𝐲 𝐭𝐡𝐢𝐬 𝐦𝐚𝐭𝐭𝐞𝐫𝐬: HyperFree achieved state-of-the-art performance on 11 datasets across classification, target detection, anomaly detection, and change detection tasks - all in a tuning-free manner. In some cases, it even outperformed specialized models that were specifically trained on each dataset. This breakthrough makes hyperspectral AI more practical and accessible, reducing the computational burden and expertise required for deployment. The work addresses a fundamental scalability challenge in Earth observation AI, potentially accelerating adoption of hyperspectral analysis across environmental monitoring, precision agriculture, and defense applications. https://lnkd.in/eKsch3eB #HyperspectralImaging #AI #MachineLearning #RemoteSensing #CVPR2025 #EarthObservation

'A satellite imaging company that played a key role in revealing Russian forces massing on Ukraine’s border prior to invasion launched a new product Wednesday that uses AI and satellite data to provide “predictive intelligence” on hundreds of sites around the world. Maxar’s new product, “Sentry”, provides a way for multiple satellite companies to collaborate and share data in order to keep more sensors on emerging developments. Maxar described Sentry as AI-powered software that can function as its own mini intelligence agency, bringing together data from not only high-resolution imaging satellites but also other intelligence sources, potentially including synthetic aperture radar satellites that use microwave pulses to “see” through clouds or at night, electro-optical satellites that can measure things like weather patterns and vegetation. Sentry can also “orchestrate” satellite data collection—meaning task multiple satellite constellations to go and collect at a specific time and place—to ensure important developments don’t go unnoticed.' https://lnkd.in/gRNvMa8c