Near-Distance Raw and Reconstructed Ground Based SAR Data

**The usage for research purposes is for free. If you use this dataset, please cite the following paper along with the dataset:
Kačan, M.; Turčinović, F.; Bojanjac, D.; Bosiljevac, M. Deep Learning Approach for Object Classification on Raw and Reconstructed GBSAR Data. Remote Sens. 2022, 14, 5673. https://doi.org/10.3390/rs14225673

The dataset consists of 2 sets: first one (RealSAR-RAW) contains raw radar data obtained using 24 GHz Ground Based Synthetic Aperture Radar (GBSAR); second one (RealSAR-IMG) contains radar images generated with reconstruction algorithm applied on that raw data.

Each example in sets represent either raw data (in RealSAR-RAW) or reconstructed image (in RealSAR-IMG) of one observed scene.
Scene can contain none, one or more test objects.
Test objects are three bottles: one aluminium, one glass, and one plastic. Bottles dimensions are the same.
Total number of scene combinations is 8 (2^3). Objects were set in various positions in near-distance range between 20 and 70 cm away from radar.
Each example is stored as matrix in .txt format (in RealSAR-RAW) and as image in .png format (in RealSAR-IMG). Matrix is size of 1024x30 (1024 frequency point, 30 GBSAR steps), and image 496x369 px.
Each set contains 337 examples: 172 of them include aluminium bottle, 172 glass bottle, 179 plastic bottle, and 29 examples are without objects.

GBSAR Specs:
Raw data is obtained using developed GBSAR system. It works in stop-and-go mode with 1 cm step size and total aperture of 30 cm. GBSAR sensor is FMCW (Frequency Modulated Continuous Wave) radar with central frequency at 24 GHz and 700 MHz wide bandwidth. FMCW radar periodically transmits sawtooth signal every 166 ms. Wave polarization of FMCW radar can be either HH or VV.