QuaSAR Compact synthetic aperture radar station for UAVs

About the Software and Hardware System

QuaSAR is a software and hardware system consisting of a compact synthetic aperture radar station and a ground control station. The ground station controls radar operating modes, receives and processes radar data, and automatically georeferences generated radar images on a geographic basemap in real time.

The system is designed to generate high-detail radar images of terrain areas regardless of illumination and atmospheric conditions, including fog, cloud cover, and precipitation.

The radar can be carried by crewed and uncrewed fixed-wing aircraft, helicopters, and multicopters with payload capacity above 2 kg.

Operating Modes

• Spotlight mode - the image is formed within the antenna beam. This provides maximum resolution, while the image coverage is limited to the area captured by the radar antenna beam;
• Stripmap mode - the image is formed continuously as the carrier moves, by stitching strips together. This mode provides wide terrain coverage throughout the route while maintaining stable detail quality;
• Video mode - information is transmitted as a stream assembled from individual spotlight radar image frames with temporal sampling: each frame is synthesized independently from a sliding time window of the hologram.

The Radar Provides:

• Radar image generation with transmission to the ground automated workstation in real time, with latency of no more than 30 s;
• Coordinate determination for detected objects;
• Automatic object detection using a pretrained convolutional neural network, including detection that accounts for typical object placement on the terrain;
• Integration with onboard radio-electronic equipment over various interfaces;
• Operation on Windows and Linux.

Features

Spotlight Images

In spotlight mode, the image is formed within the antenna beam. This provides maximum resolution, while the image coverage is limited to the area captured by the radar antenna beam.

Stripmap Images

The image is formed continuously as the carrier moves, by stitching strips together. This mode provides wide terrain coverage throughout the route while maintaining stable detail quality.

Video Stream

Video mode is implemented as frame-by-frame synthesis of spotlight radar images with temporal sampling: each frame is synthesized independently from a sliding time window of the hologram. The result is encoded in H.265/HEVC through FFmpeg with configurable quality parameters (CRF 18-28) and a frame rate of up to 120 fps.

Object Recognition

The system includes an object recognition module based on the YOLO11 OBB neural network. Unlike standard detectors, the OBB model accounts for object rotation angle, which is critical for radar imagery.

The recognition module automatically labels objects in radar images, such as equipment, vehicles, and buildings.

Supported Platforms

The software package is designed as a cross-platform solution with a focus on embedded systems and runs successfully on a wide range of devices.

NVIDIA Jetson Orin Family

This is the system’s target platform. The project is optimized for NVIDIA modules, providing maximum performance per watt when processing radar data in field conditions. Supported modules:

• Jetson Orin AGX
• Jetson Orin NX
• Jetson Orin Nano

Features: full hardware acceleration support through NVIDIA CUDA (sm_87/compute_87 architecture) and use of specialized codecs for video stream multiplexing.

Single-Board Computers (SBC)

Although these platforms lack hardware acceleration, efficient parallel CPU computation allows the software package to run on single-board computers such as Raspberry Pi 4/5.

Personal Computers

• Linux (x86_64): the primary development and analysis environment. PCs with discrete NVIDIA GPUs using Ampere or Blackwell architectures achieve the highest radar image generation speed.

Platform Comparison

The comparison uses the time required to generate a stripmap radar image of a 3,000 x 4,529 m terrain area.