Author: Aidan Rosenblatt, Gabriela Crother-Collado
This project synthesizes deep learning training data for a millimeter wave radar imaging in a similar manner to the "HawkEye" project from the research paper Through Fog High Resolution Imaging Using Millimeter Wave Radar. This synthesizer is implemented entirely in MATLAB and uses a dataset of car CAD models from the HawkEye GitHub repository.
Unlike computer-vision-based solutions, mm wave radars are able to penetrate dense fog, making them an attractive alternative solution for use in automatic driving systems.
This project relies on only MATLAB scripts/functions. All requirements listed below.
- Required Add-Ons:
- DSP System Toolbox
- Computer Vision Toolbox
- Image Processing Toolbox
- Download the previously listed add-ons from the matlab add on explorer.
- Download the "Synthesizer" folder from this repository.
- Add the "scripts" subfolder to your MATLAB path.
- Open "main.m".
- Enter the path to the CAD file from the "CAD" folder to be processed on line 4.
- Run the "main.m" script.
CAD_model_x.mat: Dataset of 36 CAD files depicting different types of cars ** For storage reasons, only two CAD files have been included in this repository. ** The other 34 CAD files in the dataset can be downloaded from this repository: https://github.com/JaydenG1019/HawkEye-Data-Code
main.m: Main training data synthesization script which calls helper functions and outputs figures containing training dataocclude_points.m: Simulated visually occluded points from radar POVshininess.m: Approximates specularity of surface points on car and outputs point cloud of reflective surface "blobs" to bounce radar beams off ofvariable_library_radar: Contains radar characteristics for simulating FMCW mm wave radar- functions folder: contains radar signal and transmission simulation functions from the HawkEye GitHub repository (not implemented by our group)
- [Through Fog High Resolution Imaging Using Millimeter Wave Radar] (https://jguan.page/HawkEye/)