sar

Keelix is a new open-source tool that uses free Sentinel-1 SAR satellite imagery to monitor naval bases without AIS data, detecting ship arrivals/departures and flagging unusual activity through before-and-after radar comparisons. Validated over 12 months at Norfolk Naval Base, it achieved ~65% detection rate with ~95% precision using signal processing rather than machine learning.

This feasibility study benchmarks a small convolutional autoencoder against block-adaptive quantization (BAQ) on AFRL GOTCHA SAR phase-history data, finding the learned coder loses at every rate. It also highlights evaluation pitfalls like CFAR border artifacts and clipping-range sensitivity, and proposes an evaluation protocol for learned radar compression.

Presents a globally trained CNN for forest above-ground biomass estimation using multi-sensor satellite data, with a sparse field calibration workflow to adapt predictions locally. Achieves improved accuracy over uncalibrated global models and ESA CCI products.

Introduces GEOID-Flood, a large-scale multi-modal benchmark dataset for flood segmentation with over 14,000 tiles from 219 events across 65 countries, evaluating foundation models against conventional encoders across single-image, multi-temporal, and multi-modal protocols.

This paper introduces a novel uncertainty-aware PINN framework for flood inference from SAR data, addressing 'physics shock' by dynamically relaxing physical constraints in noisy regions. Evaluated on Sen1Floods11, the method achieves a 25% improvement in IoU and provides calibrated uncertainty bounds for operational disaster response.