Neural network-based object surface estimation in radar system
Systems and methods to perform object surface estimation using a radar system involve receiving reflected signals resulting from reflection of transmit signals by an object. The method includes processing the reflected signals to obtain an image. The image indicates an intensity associated with at least one set of angle values and a set of range values. The method also includes processing the image to provide the object surface estimation. The object surface estimation indicates a subset of the at least one set of angle values and associated ranges within the set of range values.
1. A method to perform object surface estimation using a radar system, the method comprising:
receiving reflected signals resulting from reflection of transmit signals by an object;
processing, using a processor, the reflected signals to obtain an image, the image indicating an intensity associated with at least one set of angle values and a set of range values;
processing, using the processor, the image to provide the object surface estimation, the object surface estimation indicating a subset of the at least one set of angle values and associated ranges within the set of range values, wherein the receiving the reflected signals includes using a two-dimensional array of antenna elements, the image is a three-dimensional image indicating the intensity associated with a first set of angle values, a second set of angle values, and the set of range values providing the object surface estimation includes indicating azimuth angle values from which the reflected signals originate, associated minimum and maximum elevation angles, and associated minimum and maximum ranges.
2. The method according to claim 1 , further comprising training a neural network to implement the processing of the image.
3. The method according to claim 2 , wherein the training the neural network includes implementing a supervised learning process by calculating a loss based on an output of the neural network and on ground truth.
4. The method according to claim 3 , wherein the training the neural network includes providing the loss as feedback to the neural network.
5. The method according to claim 1 , further comprising locating the radar system in a vehicle and controlling an operation of the vehicle based on the object surface estimation.
6. A system to perform object surface estimation using a radar system, the system comprising:
a plurality of antenna elements configured to receive reflected signals resulting from reflection of transmit signals by an object; and
a processor configured to process the reflected signals to obtain an image, the image indicating an intensity associated with at least one set of angle values and a set of range values, and to process the image to provide the object surface estimation, the object surface estimation indicating a subset of the at least one set of angle values and associated ranges within the set of range values, wherein
the plurality of antenna elements is arranged as a one-dimensional array of antenna elements, the image is a two-dimensional image indicating the intensity associated with the set of angle values and the set of range values, and the object surface estimation includes an indication of azimuth angle values from which the reflected signals originate and associated minimum and maximum ranges, or
the plurality of antenna elements is arranged as a two-dimensional array of antenna elements, the image is a three-dimensional image indicating the intensity associated with a first set of angle values, a second set of angle values, and the set of range values, and the object surface estimation includes an indication of azimuth angle values from which the reflected signals originate, associated minimum and maximum elevation angles, and associated minimum and maximum ranges.
7. The system according to claim 6 , wherein the processor implements a neural network to process the image.
8. The system according to claim 7 , wherein the neural network is trained by a supervised learning process that includes calculating a loss based on an output of the neural network and on ground truth.
9. The system according to claim 8 , wherein the loss is provided as feedback to the neural network in the training.
10. The system according to claim 6 , wherein the radar system is in or on a vehicle and an operation of the vehicle is controlled based on the object surface estimation.
11. A method to perform object surface estimation using a radar system, the method comprising:
receiving reflected signals resulting from reflection of transmit signals by an object;
processing, using a processor, the reflected signals to obtain an image, the image indicating an intensity associated with at least one set of angle values and a set of range values;
processing, using the processor, the image to provide the object surface estimation, the object surface estimation indicating a subset of the at least one set of angle values and associated ranges within the set of range values, wherein the receiving the reflected signals includes using a horizontal one-dimensional array of antenna elements, the image is a two-dimensional image indicating the intensity associated with the set of angle values and the set of range values, and providing the object surface estimation includes indicating azimuth angle values from which the reflected signals originate and associated minimum and maximum ranges.