Synthetic traffic object generator

• US 10,223,601 B1
• Filed: 10/12/2017
• Issued: 03/05/2019
• Est. Priority Date: 10/12/2017
• Status: Active Grant

• Current Assignee: Denso International America Incorporated (DENSO Corporation)
• Original Assignee: Denso International America Incorporated (DENSO Corporation)
• Inventors: Huang, Minglei
• Primary Examiner(s): Park, Soo Jin
• Application Number:
  US15/782,061
• Time in Patent Office: 509 Days
• Field of Search: None
• CPC Class Codes: G06F 18/217 Validation; Performance evaluation; Active pattern learning techniques
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06F 18/2414 Smoothing the distance, e.g. radial basis function networks [RBFN]
• CPC Class Codes: G06T 2207/20081 Training; Learning
• CPC Class Codes: G06T 2207/30256 Lane; Road marking
• CPC Class Codes: G06V 10/82 using neural networks
• CPC Class Codes: G06V 20/56 exterior to a vehicle by using sensors mounted on the vehicle
• CPC Class Codes: G06V 20/582 of traffic signs
• CPC Class Codes: G06V 20/584 of vehicle lights or traffic lights
• CPC Class Codes: G06V 20/588 Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
• CPC Class Codes: G06V 30/1916 Validation; Performance evaluation
• CPC Class Codes: G06V 30/19173 Classification techniques
• CPC Class Codes: G06V 30/194 References adjustable by an adaptive method, e.g. learning

Abstract

A method for generating images of traffic objects using deep learning is disclosed. The method includes obtaining image data from an image. The method includes assigning a set of default parameters to the image data, and the set of default parameters include values associated with at least one of a weather condition of the image and a defect condition of an object of the image. The method includes generating a set of predicted parameters based on the image data. The method includes determining an error for each parameter of the set of predicted parameters, and the error is based on a value of the parameter of the set of predicted parameters and a value of a corresponding default parameter of the set of default parameters. The method includes adjusting a weight of a corresponding connection of the deep learning module.

Claims

1. A method comprising:

obtaining, using a deep learning module that is implemented by a processor configured to execute instructions stored in a non-transitory memory, image data from an image;

assigning, using the deep learning module, a set of default parameters to the image data, wherein the set of default parameters include values associated with at least one of a weather condition of the image and a defect condition of an object of the image;

generating, using the deep learning module, a set of predicted parameters based on the image data;

determining, using the deep learning module, an error for each parameter of the set of predicted parameters, wherein the error is based on a value of the parameter of the set of predicted parameters and a value of a corresponding default parameter of the set of default parameters; and

adjusting, using the deep learning module and in response to the error for one parameter of the set of predicted parameters being greater than an error threshold, a weight of a corresponding connection of the deep learning module.

2. The method of claim 1 further comprising generating, using the deep learning module, an image based on a set of test parameters.

3. The method of claim 1 further comprising identifying, using the deep learning module and in response to receiving a first image, the object of the image based on a set of object parameters associated with the first image.

4. The method of claim 3, wherein identifying the object is based on a comparison of the set of object parameters to the set of default parameters.

5. The method of claim 1, wherein the set of default parameters include parameters associated with an edge of the object, a line of the object, a position of the object, an orientation of the object, a scale of the object, an object type of the object, and a lighting condition of the image.

6. The method of claim 1, wherein generating the set of predicted parameters based on the image data includes filtering, using a filtering module that is configured to perform a convolution operation, the image data.

7. The method of claim 1, wherein generating the set of predicted parameters based on the image data includes determining, using a pooling module of the deep learning module, a maximum value from a set of the image data.

8. The method of claim 1, wherein generating the set of predicted parameters based on the image data includes converting the value of the parameter of the set of predicted parameters to zero in response to the value of the parameter of the set of predicted parameters being less than zero.

9. A system comprising:

a deep learning module that is implemented by a processor configured to execute instructions stored in a non-transitory memory, wherein the instructions include;

obtaining, using the deep learning module, image data from an image;

assigning, using the deep learning module, a set of default parameters to the image data, wherein the set of default parameters include values associated with at least one of a weather condition of the image and a defect condition of an object of the image;

generating, using the deep learning module, a set of predicted parameters based on the image data;

determining, using the deep learning module, an error for each parameter of the set of predicted parameters, wherein the error is based on a value of the parameter of the set of predicted parameters and a value of a corresponding default parameter of the set of default parameters; and

adjusting, using the deep learning module and in response to the error for one parameter of the set of predicted parameters being greater than an error threshold, a weight of a corresponding connection of the deep learning module.

10. The system of claim 9 wherein the instructions include generating, using the deep learning module, an image based on a set of test parameters.

11. The system of claim 9 wherein the instructions include identifying, using the deep learning module and in response to receiving a first image, the object of the image based on a set of object parameters associated with the first image.

12. The system of claim 11, wherein identifying the object is based on a comparison of the set of object parameters to the set of default parameters.

13. The system of claim 9, wherein the set of default parameters include parameters associated with an edge of the object, a line of the object, a position of the object, an orientation of the object, a scale of the object, an object type of the object, and a lighting condition of the image.

14. The system of claim 9, wherein generating the set of predicted parameters based on the image data includes filtering, using a filtering module that is configured to perform a convolution operation, the image data.

15. The system of claim 9, wherein generating the set of predicted parameters based on the image data includes determining, using a pooling module of the deep learning module, a maximum value from a set of the image data.

16. The system of claim 9, wherein the set of predicted parameters based on the image data includes converting the value of the parameter of the set of predicted parameters to zero in response to the value of the parameter of the set of predicted parameters being less than zero.