KR-102964047-B1 - METHOD FOR ESTIMATING POSE, COMPUTER-READABLE STORAGE MEDIUM AND COMPUTER PROGRAM FOR CONTROLLING THE HOLDER DEVICE

Abstract

The pose estimation method of an embodiment may include the steps of: generating a plurality of images including a first image of a first reference size and a second image of a second reference size larger than the size of the first image based on an input image collected from a camera; extracting feature points of the first image and the second image; generating a plurality of first pairs by matching the feature points of the first image with feature points of a DB image; estimating the pose of the first image using the feature points of the first pair among the plurality of first pairs that satisfy a first reference value; generating a plurality of second pairs by matching the feature points of the second image for the first image with the feature points of the DB image; estimating the pose of the second image using the feature points of the second pair among the plurality of second pairs that satisfy a second reference value; and determining the pose of the input image based on the estimated poses of the plurality of second images. The embodiment has the effect of estimating the pose of an image more accurately by removing mismatched feature points using pairs of feature points.

Inventors

• 임승욱
• 유연걸
• 윤찬민
• 정준영

Assignees

• 에스케이텔레콤 주식회사

Dates

Publication Date

20260511

Application Date

20191029

Claims (10)

1. A method performed in a pose estimation device for estimating the pose of a camera equipped in a terminal, A step of generating a plurality of images including a first image of a first reference size and a second image of a second reference size larger than the size of the first image, based on an input image collected from the camera; A step of extracting feature points of the first image and the second image; A step of generating a plurality of first pairs by matching feature points of the first image with feature points of the DB image; A step of estimating the pose of the first image using the feature points of the first pair satisfying a first reference value among the plurality of first pairs; A step of generating a plurality of second pairs by matching feature points of the second image with the first image in which the pose is estimated with feature points of the DB image; A step of estimating the pose of the second image using the feature points of the second pair satisfying a second reference value among the plurality of second pairs; and A step of determining the pose of the input image based on the pose of the plurality of second images estimated above, while considering the number of matched feature points between the second image and the plurality of DB images and the error rate of the average distance between the matched feature points. A pose estimation method including
2. In paragraph 1, The step of estimating the pose of the first image above is, A step of selecting one of the plurality of first pairs as a first reference pair and comparing the first reference pair with the plurality of first pairs; A step of selecting any one of the plurality of first pairs, excluding the first reference pair, as a second reference pair, and comparing the second reference pair with the plurality of first pairs; and A step of estimating the pose of the first image using the feature points of the plurality of first pairs that satisfy the first reference value for the first reference pair and the second reference pair. A pose estimation method including
3. In paragraph 2, A pose estimation method in which the first reference value is a value satisfying that the distance between the first reference pair and the distance between any one of the first pairs is less than the first distance.
4. In paragraph 2, A pose estimation method in which the first reference value is a value satisfying that the ratio of the distances of the first reference pair and the ratio of the distances between any one of the first pairs is less than the first ratio.
5. In paragraph 2, The step of estimating the pose of the first image involves selecting at least three of the plurality of remaining pairs that satisfy the first reference value, and a pose estimation method for estimating the pose of the first image.
6. In paragraph 1, The step of generating the above plurality of images is, A pose estimation method in which, if the size of the input image is larger than the second reference size, the size of the input image is changed to a size between the first image and the second image, and if the size of the input image is smaller than the second reference size, the input image is used as the second image.
7. delete
8. In paragraph 1, The step of generating the above plurality of images is, A pose estimation method for generating the first image and the second image by changing the size of the input image along the major axis.
9. As a computer-readable recording medium storing a computer program, The above computer program is, A computer-readable recording medium comprising instructions for a processor to perform a method according to any one of paragraphs 1 through 6 and paragraph 8.
10. As a computer program stored on a computer-readable recording medium, The above computer program is, A computer program comprising instructions for a processor to perform a method according to any one of paragraphs 1 through 6 and paragraph 8.

Description

Method for Estimating Pose, Computer-Readable Storage Medium and Computer Program for Controlling the Holder Device The embodiment relates to a pose estimation method for effectively estimating the pose of a camera equipped in a terminal. Augmented Reality (AR) is a technology that uses location and orientation information to determine an approximate location, identifies the services desired by the user through a comparison between facility information—such as surrounding buildings—and real-world video information input based on camera movement, and provides relevant information. More specifically, augmented reality is a field of virtual reality (VR) that uses computer graphics techniques to superimpose virtual objects onto a real environment, making them appear as if they exist within the original environment. Unlike conventional virtual reality, which focuses solely on virtual spaces and objects, augmented reality is a technology that superimposes virtual objects onto a foundation of the real world to provide supplementary information that is difficult to obtain from the real world alone. With the commercialization of 5G communication, such augmented reality technology is gaining prominence in the field of mobile AR technology used in communication terminals, and marker-based mobile AR technology or sensor-based mobile AR technology is generally used in current mobile AR technology applications. Marker-based mobile AR technology is a technology that recognizes a real object by recognizing a marker corresponding to the real object when photographing a real object to be augmented using a virtual object, and sensor-based mobile AR technology is a technology that infers the current location and the direction the terminal is looking by using GPS and a digital compass installed in the terminal, and overlays POI (Point of Interests) information corresponding to the image in the inferred direction. However, marker-based mobile AR technology has the problem that it cannot augment virtual objects without markers, and sensor-based mobile AR technology has the problem of being unable to accurately augment virtual objects onto specific objects due to errors in the current location and orientation of the detected device. FIG. 1 is a block diagram schematically showing an apparatus for performing a pose estimation method according to an embodiment. FIG. 2 is a flowchart illustrating a pose estimation method according to an embodiment. FIG. 3 is a diagram illustrating the step of generating an image according to an embodiment. FIG. 4 is a diagram illustrating the step of extracting feature points of an image according to an embodiment. FIG. 5 is a diagram illustrating the step of generating a pair according to an embodiment. FIG. 6 is a flowchart illustrating the step of estimating a first pose according to an embodiment. FIGS. 7 to 11 are drawings for explaining the step of estimating a first pose according to an embodiment. Hereinafter, embodiments will be described in detail with reference to the drawings. FIG. 1 is a block diagram schematically showing an apparatus for performing a pose estimation method according to an embodiment. Referring to FIG. 1, the pose estimation method according to the embodiment can be performed in a pose estimation device (100). The pose estimation device (100) may receive an input image from a terminal (200). The terminal (200) may include various terminals such as mobile phones and computers. The terminal (200) may be equipped with a camera or a sensor. The pose estimation device (100) may obtain an input image from the camera or sensor of the terminal (200). The pose estimation device (100) can receive DB images from DB (300). DB (300) may be an area where accurate image information is stored. DB (300) may be created internally by collecting actual images and may be an area of a portal site such as Naver, Daum, or Google. The pose estimation device (100) can estimate the position of the terminal (200) using the input image and DB image, and can establish a foundation for providing AR services using this. The pose estimation device (100) can generate input images into multiple images. The pose estimation device (100) can generate an input image having a first image having a first reference size and a second image having a second reference size larger than the first reference size. Alternatively, the pose estimation device (100) may generate three or more images. The sizes of the first and second images may be changed to be smaller than the size of the input image. The second image may be the same size as the input image. The pose estimation device (100) can extract feature points for a first image. The pose estimation device (100) can extract feature points for a second image. Since the size of the first image is smaller than the size of the second image, the number of feature points in the first image may be less than the number of feature points in the second image. The pose estimatio