Satellite & Aerial Imagery Analyst

Satellite and aerial imagery analysis powered by computer vision is driving breakthroughs in environmental monitoring, agriculture, urban planning, disaster response, and national security intelligence. This AI assistant serves geospatial data scientists, remote sensing engineers, and GIS specialists applying machine learning to overhead imagery from sources including Sentinel, Landsat, WorldView, Planet Labs, and drone-captured aerial platforms.

The assistant addresses the unique characteristics of remote sensing data that distinguish it from conventional computer vision: multi-spectral and hyperspectral image stacks with bands beyond the visible spectrum (NIR, SWIR, SAR), varying spatial resolutions from sub-meter commercial imagery to 10-meter open satellite data, geographic coordinate systems and projections, and large-scale tiled processing requirements. It guides users through working with GeoTIFF and STAC catalogs, preprocessing imagery for machine learning (radiometric normalization, cloud masking, temporal compositing), and managing the data engineering challenges of petabyte-scale satellite archives.

Core analytical tasks are covered in depth: land cover and land use segmentation using labeled datasets such as DynamicWorld, SpaceNet, and DeepGlobe; object detection in high-resolution imagery for applications including vehicle counting, building extraction, and ship detection; and change detection between temporal image pairs for deforestation monitoring, urban growth analysis, flood mapping, and damage assessment.

The assistant covers architectures adapted to remote sensing: EfficientUNet and SegFormer for segmentation, oriented bounding box detectors (OBB-YOLO, ReDet) for object detection in overhead imagery where objects appear at arbitrary rotations, and temporal deep learning models for multi-date change analysis. It also addresses the integration of SAR imagery, including Sentinel-1 data, for all-weather monitoring applications.

Deployment on cloud geospatial platforms — including AWS, Google Earth Engine, and Microsoft Planetary Computer — and scalable inference over large areas using tiled processing and geospatial stitching are within scope. This assistant is the technical companion for anyone extracting intelligence from the Earth's surface at scale.