Method and apparatus for recognizing object

Patent metadata

Field	Value
Publication number	US-11048266-B2
Application number	US-201815992447-A
Country	US
Kind code	B2
Filing date	May 30, 2018
Priority date	Sep 4, 2017
Publication date	Jun 29, 2021
Grant date	Jun 29, 2021

Abstract

A method and apparatus for recognizing an object are provided, the method including extracting a feature from an input image and generating a feature map in a neural network. In parallel with the generating of the feature map, a region of interest (ROI) corresponding to an object of interest is extracted from the input image, and a number of object candidate regions used to detect the object of interest is determined based on a size of the ROI. The object of interest is recognized from the ROI based on the number of object candidate regions in the neural network.

First claim

What is claimed is:

1. An object recognition method comprising:
extracting, in a neural network, a feature from an input image and generating a feature map;
extracting, in parallel with the generating of the feature map, a region of interest (ROI) corresponding to an object of interest from the input image;
determining a number of object candidate regions used to detect the object of interest based on a size of the ROI and a size of the input image; and
recognizing the object of interest from the ROI based on the number of object candidate regions in the neural network.

2. The object recognition method of claim 1 , wherein the object of interest comprises any one or any combination of a road, a vehicle, a human, an animal, a plant, and a building.

3. The object recognition method of claim 1 , wherein the determining of the number of object candidate regions comprises:
calculating a ratio of the size of the ROI to the size of the input image; and
determining the number of object candidate regions based on the ratio.

4. The object recognition method of claim 1 , wherein the recognizing of the object of interest comprises:
determining positions of the object candidate regions on the feature map; and
recognizing the object of interest from the ROI based on the positions of the object candidate regions.

5. The object recognition method of claim 1 , wherein the extracting of the ROI comprises extracting the ROI based on any one or any combination of a training-based scene segmentation algorithm and an image processing algorithm.

6. The object recognition method of claim 1 , wherein the neural network comprises a region-based convolutional neural network (R-CNN) comprising a region proposal network (RPN) and a detection network.

7. The object recognition method of claim 1 , further comprising:
determining a control parameter to control a speed of a vehicle and a traveling direction of the vehicle based on a result of the recognizing; and
controlling a movement of the vehicle using the control parameter.

8. The object recognition method of claim 1 , wherein the determining of the number of the object candidate regions comprises:
calculating a ratio of the size of the ROI to the size of the input image; and
determining the number of object candidate regions based on applying a number of default object candidate regions for the neural network to the ratio.

9. The object recognition method of claim 1 , wherein the ROI comprises a region corresponding to one or any combination of a road, a vehicle, a human, an animal, a plant, and a building.

10. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, cause the processor to perform the method of claim 1 .

11. An object recognition method comprising:
extracting a region of interest (ROI) corresponding to an object of interest from an input image;
determining, based on a size of the ROI and a size of the input image, a number of object candidate regions used to detect the object of interest; and
recognizing, in a neural network, the object of interest from the ROI based on the number of object candidate regions.

12. The object recognition method of claim 11 , wherein the object of interest comprises any one or any combination of a road, a vehicle, a human, an animal, a plant and a building.

13. The object recognition method of claim 11 , wherein the determining of the number of object candidate regions comprises:
calculating a ratio of the size of the ROI to the size of the input image; and
determining the number of object candidate regions based on the ratio.

14. The object recognition method of claim 11 , wherein the recognizing of the object of interest comprises:
determining positions of the object candidate regions on a feature map generated in the neural network, based on the number of object candidate regions; and
recognizing the object of interest from the ROI based on the positions of the object candidate regions.

15. The object recognition method of claim 11 , wherein the extracting of the ROI comprises extracting the ROI based on any one or any combination of a training-based scene segmentation algorithm and an image processing algorithm.

16. The object recognition method of claim 11 , wherein the neural network comprises a region-based convolutional neural network (R-CNN) comprising a region proposal network (RPN) and a detection network.

17. The object recognition method of claim 11 , further comprising:
determining a control parameter used to control a speed of a vehicle and a traveling direction of the vehicle based on a result of the recognizing; and
controlling a movement of the vehicle using the control parameter.

18. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, cause the processor to perform the method of claim 11 .

19. An object recognition apparatus comprising:
a sensor configured to acquire an input image;
a neural network, comprising a plurality of layers, configured to extract a feature from the input image and to generate a feature map; and
a processor configured to extract, in parallel with a generation of the feature map, a region of interest (ROI) corresponding to an object of interest from the input image, and to determine, based on a size of the ROI and a size of the input image, a number of object candidate regions used to detect the object of interest,
wherein the neural network is further configured to recognize the object of interest from the ROI based on the number of object candidate regions.

20. The object recognition apparatus of claim 19 , wherein the processor is further configured to calculate a ratio of the size of the ROI to the size of the input image and to determine the number of object candidate regions based on the ratio.

21. An object recognition apparatus comprising:
a sensor configured to acquire an input image;
a processor configured to extract a region of interest (ROI) corresponding to an object of interest from the input image and to determine, based on a size of the ROI and a size of the input image, a number of object candidate regions used to detect the object of interest; and
a neural network, comprising a plurality of layers, configured to recognize the object of interest from the ROI based on the number of object candidate regions.

22. An object recognition apparatus comprising:
a sensor configured to capture an image; and
a processor configured to
extract a region of interest (ROI) corresponding to an object of interest from the image,
calculate a ratio of a size of the ROI to a size of the image,
determine a number of object candidate regions used to detect the object of interest based on the ratio; and

a neural network configured to
extract a feature from the image and to generate a feature map, and
recognize the object of interest from the ROI based on the number of object candidate regions and the feature map.

23. The object recognition method of claim 22 , wherein the processor is further configured to determine a control parameter to control a speed of the vehicle and a traveling direction of the vehicle based on the recognized object.

Assignees

• Samsung Electronics Co Ltd

Inventors

• Jeon Paulbarom

Classifications

• G06V10/82

  using neural networks · CPC title

• G06V20/56 Primary

  exterior to a vehicle by using sensors mounted on the vehicle · CPC title

• G06F18/241

  relating to the classification model, e.g. parametric or non-parametric approaches · CPC title

• G06F18/2163

  Partitioning the feature space · CPC title

• G06N3/045

  Combinations of networks · CPC title