Granted Patent B2

US 12,249,164 · App. 17/701,727 · Granted Mar 11, 2025

Inventors: Yuji Yasui (Wako, JP); Hideki Matsunaga (Wako, JP)

Assignee: HONDA MOTOR CO., LTD.

G06V20/588B60W40/06G06F18/214B60W2420/403B60W2552/20

Abstract

Claims (58)

1. A training device comprising:

a storage device storing a program; and

a hardware processor,

wherein the hardware processor executes the program stored in the storage device to:

acquire a captured image of a road which is captured by a camera mounted on a mobile object;

estimate a pitching amount of the mobile object at each time of image capturing based on the captured image;

generate an actually captured image by performing first correction for canceling the pitching amount on the captured image;

add a computer graphic image of an object present on a road to the actually captured image based on the captured image;

generate a training image by performing second correction for undoing the first correction on the actually captured image to which the computer graphic image is added; and

train parameters of a model to output a type of the object when an image is input using the training image as training data and using a type of the added computer graphic image as teacher data.

2. The training device according to claim 1 ,

wherein the hardware processor estimates a solar radiation direction based on the actually captured image based on the captured image and gives a shadow based on the solar radiation direction to a computer graphic image of the object.

3. The training device according to claim 1 ,

wherein the captured image is a captured image captured by the camera mounted on the mobile object, and

the hardware processor acquires a movement amount of the mobile object, and determines a position and a size of the computer graphic image based on the movement amount of the mobile object.

4. An object detection device is an object detection device mounted on the mobile object, and, by inputting a captured image of at least a road in a traveling direction of the mobile object, captured by the camera mounted on the mobile object, to the trained model trained by the training device according to any one of claims claim 1 , discriminates whether an object on the road reflected in the captured image is an object with which the mobile object needs to avoid contact.

5. The training device according to claim 1 ,

wherein the hardware processor compares a first captured image at a first image capturing time with a second captured image at a second image capturing time before the first image capturing time in each time of image capturing, and estimates the pitching amount of the mobile object between the first image capturing time and the second image capturing time as the pitching amount at the first image capturing time.

6. The training device according to claim 1 ,

wherein the hardware processor determines a correction amount of the first correction using a table or a map that defines how much each pixel on a captured image vertically fluctuates with respect to the pitching amount.

7. A training method executed using a computer, comprising:

acquiring a captured image of a road which is captured by a camera mounted on a mobile object;

estimating a pitching amount of the mobile object at each time of image capturing based on the captured image;

generating an actually captured image by performing first correction for canceling the pitching amount on the captured image;

adding a computer graphic image of an object present on a road to the actually captured image based on the captured image;

generating a training image by performing second correction for undoing the first correction on the actually captured image to which the computer graphic image is added; and

training parameters of a model to output a type of the object when an image is input by using the training image as training data and using a type of the added computer graphic image as teacher data.

8. A computer-readable non-transitory storage medium that has stored a program causing a computer to:

acquire a captured image of a road which is captured by a camera mounted on a mobile object;

estimate a pitching amount of the mobile object at each time of image capturing based on the captured image;

generate an actually captured image by performing first correction for canceling the pitching amount on the captured image;

add a computer graphic image of an object present on a road to the actually captured image based on the captured image;

generate a training image by performing second correction for undoing the first correction on the actually captured image to which the computer graphic image is added; and

train parameters of a model to output a type of the object when an image is input by using the training image as training data and using a type of the added computer graphic image as teacher data.

9. A training device comprising:

a storage device that has storing a program; and

a hardware processor,

wherein the hardware processor executes the program stored in the storage device to:

acquire a captured image of a road which is captured by a camera mounted on a mobile object:

estimate a pitching amount of the mobile object at each time of image capturing based on the captured image;

generate an actually captured image by performing first correction for canceling the pitching amount on the captured image;

add a computer graphic image of an object present on a road to the actually captured image; based on the captured image,

generate a training image by performing second correction for undoing the first correction on the actually captured image to which the computer graphic image is added; and

train parameters of a model to output a position of the object when an image is input by using the training image as training data and using a position of the added computer graphic image as teacher data.

10. A training method executed using a computer, comprising:

acquiring a captured image of a road which is captured by a camera mounted on a mobile object:

estimating a pitching amount of the mobile object at each time of image capturing based on the captured image;

generating an actually captured image by performing first correction for canceling the pitching amount on the captured image;

adding a computer graphic image of an object present on a road to the actually captured image; based on the captured image,

generating a training image by performing second correction for undoing the first correction on the actually captured image to which the computer graphic image is added; and

training parameters of a model to output a position of the object when an image is input by using the training image as training data and using a position of the added computer graphic image as teacher data.

11. A computer-readable non-transitory storage medium that has stored a program causing a computer to:

acquire acquiring a captured image of a road which is captured by a camera mounted on a mobile object;

estimate a pitching amount of the mobile object at each time of image capturing based on the captured image;

generate an actually captured image by performing first correction for canceling the pitching amount on the captured image;

add adding a computer graphic image of an object present on a road to the actually captured image based on the captured image;

generate a training image by performing second correction for undoing the first correction on the actually captured image to which the computer graphic image is added; and

train training parameters of a model to output a position of the object when an image is input by using the training image as training data and using a position of the added computer graphic image as teacher data.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (23)