HR-SAR-Net: A Deep Neural Network for Urban Scene Segmentation from High-Resolution SAR Data
Preprint
Abstract
Synthetic aperture radar (SAR) data is becoming increasingly available to a wide range of users through commercial service providers with resolutions reaching 0.5m/px. Segmenting SAR data still requires skilled personnel, limiting the potential for large-scale use. We show that it is possible to automatically and reliably perform urban scene segmentation from next-gen resolution SAR data (0.15m/px) using deep neural networks (DNNs), achieving a pixel accuracy of 95.19% and a mean IoU of 74.67% with data collected over a region of merely 2.2km${}^2$. The presented DNN is not only effective, but is very small with only 63k parameters and computationally simple enough to achieve a throughput of around 500Mpx/s using a single GPU. We further identify that additional SAR receive antennas and data from multiple flights massively improve the segmentation accuracy. We describe a procedure for generating a high-quality segmentation ground truth from multiple inaccurate building and road annotations, which has been crucial to achieving these segmentation results.Keywords
All Related Versions
- Version 1, 2019-12-10, ArXiv
- Version 2, 2023-01-16, ArXiv
- Published version: , 1.
This publication has 14 references indexed in Scilit:
- A Back-Projection Tomographic Framework for VHR SAR Image Change DetectionIEEE Transactions on Geoscience and Remote Sensing, 2019
- PolSAR Image Semantic Segmentation Based on Deep Transfer Learning—Realizing Smooth Classification With Small Training SetsIEEE Geoscience and Remote Sensing Letters, 2019
- Synthetic Aperture Radar Image Generation With Deep Generative ModelsIEEE Geoscience and Remote Sensing Letters, 2018
- Buildings Detection in VHR SAR Images Using Fully Convolution Neural NetworksIEEE Transactions on Geoscience and Remote Sensing, 2018
- Road Segmentation in SAR Satellite Images With Deep Fully Convolutional Neural NetworksIEEE Geoscience and Remote Sensing Letters, 2018
- High-Resolution PolSAR Scene Classification With Pretrained Deep Convnets and Manifold Polarimetric ParametersIEEE Transactions on Geoscience and Remote Sensing, 2018
- Adaptive Hierarchical Multinomial Latent Model With Hybrid Kernel Function for SAR Image Semantic SegmentationIEEE Transactions on Geoscience and Remote Sensing, 2018
- A Multisquint Framework for Change Detection in High-Resolution Multitemporal SAR ImagesIEEE Transactions on Geoscience and Remote Sensing, 2018
- OpenSARShip: A Dataset Dedicated to Sentinel-1 Ship InterpretationIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2017
- New global forest/non-forest maps from ALOS PALSAR data (2007–2010)Remote Sensing of Environment, 2014