System and method for generating a radar image of a scene
The present disclosure provides a system and a method for generating a radar image of a scene. The method comprises receiving radar measurements of a scene collected from a set of antennas, wherein the set of antennas are under uncertainties caused by one or a combination of position ambiguities and clock ambiguities of each of the antennas. The method further comprises generating the radar image of the scene by solving a sparse recovery problem. The sparse recovery problem determines, until a termination condition is met, a set of image shifts of the radar image corresponding to different uncertainties of the antennas and updates an estimate of the radar image, based on the determined set of image shifts of the radar image. The sparse recovery problem is solved with a neural network denoiser that denoises a filtering of the estimate of the radar image.
1. A system for generating a radar image of a scene, the system comprising: at least one processor; and a memory having instructions stored thereon that, when executed by the at least one processor, cause the system to:
receive radar measurements of the scene collected from a set of antennas, wherein the radar measurements are measurements associated with reflections of a radar pulse transmitted to the scene;
generate the radar image of the scene by iteratively solving a sparse recovery problem until a termination condition is met, wherein the sparse recovery problem is solved as a regularized optimization with a regularizer implemented using a neural network denoiser, wherein an iteration of the regularized optimization is configured to:
determine a set of image shifts of the radar image corresponding to different uncertainties of the set of antennas; and
update an estimate of the radar image, based on the determined set of image shifts of the radar image, such that, for each of the antennas, the estimate of the radar image shifted by the corresponding shift of the radar image fits the radar measurements of the antenna;
filter the estimate of the radar image by applying a forward radar operator and an adjoint radar operator to the estimate of the radar image; and
denoise the filtered estimate of the radar image using the neural network denoiser; and
render the radar image when the termination condition is met.
2. The system of claim 1 , wherein the neural network denoiser comprises a residual Unet architecture.
3. The system of claim 1 , wherein the neural network denoiser is trained based on a training dataset including input images and output images, wherein the input images include back projected images, and the output images include ground truth object radar images.
4. The system of claim 1 , wherein the scene includes at least one moving object.
5. The system of claim 4 , further comprising a tracker, wherein the tracker is configured to determine a state of the at least one moving object, based on at least the radar image.
6. The system of claim 5 , wherein the state of at least one moving object includes one or combination of a size, an orientation, and a position of the at least one moving object.
7. The system of claim 1 , wherein the set of antennas are under uncertainties caused by one or a combination of position ambiguities and clock ambiguities of each of the antennas.
8. A method for generating a radar image of a scene, comprising:
receiving radar measurements of the scene collected from a set of antennas, wherein the radar measurements are measurements associated with reflections of a radar pulse transmitted to the scene;
generating the radar image of the scene by iteratively solving a sparse recovery problem until a termination condition is met, wherein the sparse recovery problem is solved as a regularized optimization with a regularizer implemented using a neural network denoiser, wherein an iteration of the regularized optimization is configured to:
determine a set of image shifts of the radar image corresponding to different uncertainties of the set of antennas;
update an estimate of the radar image, based on the determined set of image shifts of the radar image, such that, for each of the antennas, the estimate of the radar image shifted by the corresponding shift of the radar image fits the radar measurements of the antenna;
filter the estimate of the radar image by applying a forward radar operator and an adjoint radar operator to the estimate of the radar image; and
denoise the filtered estimate of the radar image using the neural network denoiser; and
rendering the radar image when the termination condition is met.
9. The method of claim 8 , wherein the neural network denoiser includes a residual Unet architecture.
10. The method of claim 8 , wherein the neural network denoiser is trained based on a training dataset including input images and output images, wherein the input images include back projected images, and the output images include ground truth object radar images.
11. The method of claim 8 , wherein the scene includes moving objects.
12. The method of claim 8 , wherein the set of antennas are under uncertainties caused by one or a combination of position ambiguities and clock ambiguities of each of the antennas.
13. A non-transitory computer-readable storage medium embodied thereon a program executable by a processor for performing a method for generating a radar image of a scene, the method comprising:
receiving radar measurements of the scene collected from a set of antennas, wherein the radar measurements are measurements associated with reflections of a radar pulse transmitted to the scene;
generating the radar image of the scene by iteratively solving a sparse recovery problem until a termination condition is met, wherein the sparse recovery problem is solved as a regularized optimization with a regularizer implemented using a neural network denoiser, wherein an iteration of the regularized optimization is configured to:
determine a set of image shifts of the radar image corresponding to different uncertainties of the set of antennas;
update an estimate of the radar image, based on the determined set of image shifts of the radar image, such that, for each of the antennas, the estimate of the radar image shifted by the corresponding shift of the radar image fits the radar measurements of the antenna;
filter the estimate of the radar image by applying a forward radar operator and an adjoint radar operator to the estimate of the radar image; and
denoise the filtered estimate of the radar image using the neural network denoiser; and
rendering the radar image when the termination condition is met.
14. The non-transitory computer-readable storage medium of claim 13 , wherein the neural network denoiser includes a residual Unet architecture.
15. The non-transitory computer-readable storage medium of claim 13 , wherein the neural network denoiser is trained based on a training dataset including input images and output images, wherein the input images include back projected images and the output images include ground truth object radar images.
16. The non-transitory computer-readable storage medium of claim 13 , wherein the scene includes moving objects.
17. The non-transitory computer-readable storage medium of claim 13 , wherein the set of antennas are under uncertainties caused by one or a combination of position ambiguities and clock ambiguities of each of the antennas.