SYSTEMS AND METHODS FOR RECOGNIZING OBJECTS IN RADAR IMAGERY

• US 20160019458A1
• Filed: 07/08/2015
• Published: 01/21/2016
• Est. Priority Date: 07/16/2014
• Status: Active Grant

• Current Assignee: General Dynamics Mission Systems Inc. (General Dynamics Corporation)
• Original Assignee: Deep Learning Analytics, LLC (General Dynamics Corporation)
• Inventors: KAUFHOLD, John Patrick
• Granted Patent: US 9,978,013 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G01S 13/90 using synthetic aperture techniques , e.g. synthetic aperture radar [SAR] techniques
• CPC Class Codes: G01S 13/9027 Pattern recognition for feature extraction
• CPC Class Codes: G01S 13/9029 specially adapted for moving target detection within a single SAR image or within multiple SAR images taken at the same time
• CPC Class Codes: G01S 13/904 SAR modes
• CPC Class Codes: G01S 7/417 involving the use of neural networks
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/0463 Neocognitrons

Abstract

The present invention is directed to systems and methods for detecting objects in a radar image stream. Embodiments of the invention can receive a data stream from radar sensors and use a deep neural network to convert the received data stream into a set of semantic labels, where each semantic label corresponds to an object in the radar data stream that the deep neural network has identified. Processing units running the deep neural network may be collocated onboard an airborne vehicle along with the radar sensor(s). The processing units can be configured with powerful, high-speed graphics processing units or field-programmable gate arrays that are low in size, weight, and power requirements. Embodiments of the invention are also directed to providing innovative advances to object recognition training systems that utilize a detector and an object recognition cascade to analyze radar image streams in real time. The object recognition cascade can comprise at least one recognizer that receives a non-background stream of image patches from a detector and automatically assigns one or more semantic labels to each non-background image patch. In some embodiments, a separate recognizer for the background analysis of patches may also be incorporated. There may be multiple detectors and multiple recognizers, depending on the design of the cascade. Embodiments of the invention also include novel methods to tailor deep neural network algorithms to successfully process radar imagery, utilizing techniques such as normalization, sampling, data augmentation, foveation, cascade architectures, and label harmonization.

Claims

1. A system for detecting objects in a radar image stream, comprising:

a flying vehicle;

a radar sensor on the flying vehicle;

a processor on the flying vehicle;

an object detection module loaded on the processor, the object detection module comprising a deep learning algorithm trained on a plurality of labeled radar image chips;

the processor configured to receive a stream of radar image data from the radar sensor, the processor further configured to provide the received stream of radar image data to the object detection module;

the object detection module programmed to invoke the deep learning algorithm to process the received stream of radar image data into a semantic label corresponding to a detected object in the stream; and

the processor further configured to provide the semantic label to an operational control module onboard the flying vehicle.

2. The system of claim 1, wherein the radar sensor is a synthetic aperture radar sensor.

3. The system of claim 1, wherein the processor is a high-speed graphics processing unit.

4. The system of claim 1, wherein the processor is a field-programmable gate array.

5. The system of claim 1, wherein the processor is an Nvidia Tegra processor.

6. The system of claim 1, wherein the processor is further configured to normalize the received stream of radar image data.

7. A method for processing a radar data stream, comprising:

receiving, at a graphics processing unit onboard a flying vehicle, a stream of radar data generated by a synthetic aperture radar sensor;

processing, by an object recognition module on the graphics processing unit, the stream of radar data into a semantic label corresponding to an object in the stream, the object recognition module comprising a deep learning algorithm; and

providing the semantic label to an operational control module onboard the flying vehicle.

8. The method of claim 7, further comprising:

normalizing the received stream of radar data.

9. The method of claim 7, wherein the deep learning algorithm further comprises an object recognition cascade.

10. The method of claim 9, wherein the object recognition cascade comprises a neural net with more than three layers.

11. The method of claim 10, wherein one of the layers comprises a convolutional layer.

12. The method of claim 10, wherein one of the layers comprises a locally connected layer.

13. A system for assembling a database of labeled radar images for object recognition, comprising:

a processor onboard a flying vehicle, the processor configured to receive a stream of sensor data obtained from a synthetic aperture radar device onboard the flying vehicle;

the processor further configured to process the stream of sensor data into a plurality of two-dimensional radar image chips, each chip including a set of pixels comprising magnitude and phase information derived from radar return signals in the stream of sensor data;

the processor further configured to wrangle at least one of a set of semantic labels from each of the plurality of two-dimensional radar image chips; and

the processor further configured to generate a database of labeled radar image chips as a result of the wrangling.

14. A system for training a deep learning network to recognize objects in a radar image, comprising:

a processor configured to receive a database of labeled radar image chips;

the processor further configured to normalize the labeled radar image chips;

the processor further configured to train a deep learning network recognizer cascade based on the normalized and labeled radar image chips, the deep learning network recognizer cascade including a deep learning network comprising at least 5 layers;

the processor additionally configured to output the trained deep learning object recognizer cascade as a result of the training.

15. The system of claim 14, wherein at least one of the layers comprises a convolutional layer.

16. The system of claim 14, wherein at least one of the layers comprises a locally connected layer.