Object detection method, electronic apparatus and object detection system

1. An object detection method, adapted to an electronic apparatus, and the method comprises:

obtaining a first image, wherein the first image is an image to be detected;

performing a geometric transformation operation on the first image to obtain at least one second image;

combining the first image and the at least one second image to generate a combination image; and

inputting the combination image including the first image and the at least one second image into a trained deep learning model to detect a target object in the first image.

2. The object detection method according to claim 1 , wherein the step of performing the geometric transformation operation on the first image to obtain the at least one second image comprises:

rotating the first image respectively according to at least one preset angle to obtain the at least one second image.

3. The object detection method according to claim 2 , wherein the at least one second image comprises a first rotated image, a second rotated image and a third rotated image, and the first image, the first rotated image, the second rotated image and the third rotated image are not overlapping with one another and are disposed on four different positions on the combination image.

4. The object detection method according to claim 1 , wherein the step of combining the first image and the at least one second image to generate the combination image comprises:

inserting a partitioning image block between the first image and the at least one second image; and

adjusting an image dimension of the combination image.

5. The object detection method according to claim 4 , wherein the partitioning image block comprises a first color block, a second color block or a combination thereof.

6. The object detection method according to claim 5 , wherein the first color block is a black block.

7. The object detection method according to claim 1 , the method further comprising:

obtaining a raw image from a video stream; and

extracting the first image from the raw image according to a region of interest.

8. The object detection method according to claim 1 , the method further comprising:

displaying, by a display, the combination image including object information, wherein the object information comprises a bonding box used to enclose the detected target object.

9. The object detection method according to claim 1 , the method further comprising:

obtaining a plurality of raw training images in a training data set;

tagging answer objects in the raw training images;

performing a data augmentation operation on the raw training images to obtain a plurality of actual training images; and

performing a deep learning according to the actual training images and the tagged answer objects to generate the trained deep learning model, wherein the data augmentation operation is independent of the geometric transformation operation.

10. The object detection method according to claim 1 , wherein the deep learning model is a convolution neural network (CNN) model.

11. An electronic apparatus, comprising:

a storage device; and

a processor, coupled to the storage device, and configured to execute commands stored in the storage device to:

obtain a first image, wherein the first image is an image to be detected;

perform a geometric transformation operation on the first image to obtain at least one second image;

combine the first image and the at least one second image to generate a combination image; and

input the combination image including the first image and the at least one second image into a trained deep learning model to detect a target object ill the first image.

12. The electronic apparatus according to claim 11 , wherein the processor is further configured to:

rotate the first image respectively according to at least one preset angle to obtain the at least one second image.

13. The electronic apparatus according to claim 12 , wherein the at least one second image comprises a first rotated image, a second rotated image and a third rotated image, and the first image, the first rotated image, the second rotated image and the third rotated image are not overlapping with one another and are disposed on four different positions on the combination image.

14. The electronic apparatus according to claim 11 , wherein the processor is further configured to:

insert a partitioning image block between the first image and the at least one second image; and

adjust an image dimension of the combination image.

15. The electronic apparatus according to claim 14 , wherein the partitioning image block comprises a first color block, a second color block or a combination thereof.

16. The electronic apparatus according to claim 15 , wherein the first color block is a black block.

17. The electronic apparatus according to claim 11 , wherein the processor is configured to:

obtain a raw image from a video stream; and

extract the first image from the raw image according to a region of interest.

18. The electronic apparatus according to claim 11 , wherein the electronic apparatus further comprises: a display, displaying the combination image including object information, wherein the object information comprises a bonding box used to enclose the detected target object.

19. The electronic apparatus according to claim 11 , wherein the processor is configured to:

obtain a plurality of raw training images in a training data set;

tag answer objects in the raw training images;

perform a data augmentation operation on the raw training images to obtain a plurality of actual training images; and

perform a deep learning according to the actual training images and the tagged answer objects to generate the trained deep learning model, wherein the data augmentation operation is independent of the geometric transformation operation.

20. An object detection system, comprising:

an image capturing device, configured to capture images; and

an electronic apparatus, connected to the image capturing device, and comprising:

a storage device; and

a processor, coupled to the storage device, and configured to execute commands stored in the storage device to:

obtain a first image, wherein the first image is an image to be detected;

perform a geometric transformation operation on the first image to obtain at least one second image;

combine the first image and the at least one second image to generate a combination image; and

input the combination image including the first image and the at least one second image into a trained deep learning model to detect a target object in the first image,

wherein the processor is further configured to:

rotate the first image respectively according to at least one preset angle to obtain the at least one second image, wherein the at least one second image comprises a first rotated image, a second rotated image and a third rotated image, and the first image, the first rotated image, the second rotated image and the third rotated image are disposed on four corners on the combination image.