KR101556937B1 - Augmented Reality Image Recognition System Using Overlap Cut Image and Method Thereof - Google Patents

Abstract

The present invention relates to a method and apparatus for effectively reducing a recognition error due to a similar image and effectively improving a user's immersion feeling and learning effect by using an overlap-cut image recognition structure for combining 3D digital contents for a printed matter, A video device provided for inputting and outputting an image; A reference image storage unit for sorting the feature regions of the image into overlapping image elements and constructing and storing the image regions by using an overlap-cut method; An image recognition unit for generating a histogram from image elements input from the image device and determining a similar image among the reference images stored in the reference image storage unit as a recognition image and enhancing the recognition image into an image or a 3D object model; Wherein the image recognition unit includes a primary recognition unit for performing a determination as a recognized image immediately when there is no similar or confused reference image in the recognition process for the image, When at least two images are recognized, the image element histogram of the image image is compared with the image element information of the reference images stored in the reference image storage unit, and the most similar reference image is searched to determine the recognition image 2 to perform identification It provides an augmented reality image recognition system using the overlapped-cut image including recognition unit.

Description

Technical Field [0001] The present invention relates to an Augmented Reality Image Recognition System using an overlap-cut image,

The present invention relates to an augmented reality image recognition system using an overlap cut image and a recognition method thereof, and more particularly, to an improved augmented reality image recognition system using an overlap cut image recognition structure for combining 3D digital contents for printed matter, To an ARP image recognition system using an overlap-cut image which can effectively improve a user's immersion feeling and a learning effect, and a recognition method thereof.

Generally, augmented reality (AR) is a term derived from virtual environments and virtual reality, which means a virtual reality in which a user views a real environment viewed by the user and a virtual environment having additional information together as a single image. Recently, with the development of science and technology, it was developed to be systematized as a structure that can use three-dimensional space in advanced equipment such as computer or project. In other words, the augmented reality system combines the real world and the virtual world so that they can interact with the user in real time.

In order to implement the augmented reality, a predetermined reference image is recognized along with an image input from an input device such as a camera, and a process of enhancing a virtual object associated with the reference image is performed.

As a conventional technique disclosed in the above-described augmented reality system, Korean Registered Patent No. 1086068 (November 16, 2011) discloses an augmented reality and mixed reality providing apparatus using a three-dimensional marker recognition method, An image data collection device for collecting image data of a real image to be imaged; An input image processing apparatus for recognizing and tracking a marker to be used as a virtual entity from the collected image data to extract a marker ID to generate virtual entity information; An output image processing device for processing and outputting the virtual entity information generated through the image data collection device and the input image processing device so as to be projected and synthesized on the real image data; And a display device for receiving and displaying the image data output from the output image processing device. Accordingly, an augmented reality and a mixed reality providing device using a three-dimensional marker recognition method capable of realizing an efficient augmented reality and a mixed reality system Is known.

However, in order to realize the augmented reality from the known art, since it is affected by the quality of the input image (noise, contrast, distortion, etc.) or the characteristics of the images to be recognized, . In other words, in the case of a book such as a fairy tale book, images are mostly composed of simple images, and it is difficult to recognize images due to the characteristics of recognition images.

Further, in order to recognize an image, unique characteristic parts (feature points) of the image alone are extracted and compared with each other. In the image having a simple change, there are not many unique feature points, there was.

Also, since each page of an infant book is often composed of similar images (e.g., characters, etc.), it is more difficult to recognize the images since the recognition mistakes are not easy to identify because of the increase of the mistakes. In other words, when we want to recognize images with many similar features, we extract feature points for each reference image at a time and compare them. Therefore, when the other parts are relatively small compared to similar parts, There is a problem in that it is considerably deteriorated.

Furthermore, if the number of images to be identified increases, a significant speed reduction occurs in a mobile device having a limited computing capability, thereby causing difficulties in real-time augmented reality application.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and it is an object of the present invention to provide a method and system for recognizing images of a plurality of types, A real-time image recognition system and a recognition method thereof.

In addition, since the present invention is configured to divide an area of an image by an overlapping method and to divide a variable area and a number area into a plurality of areas, it is possible to reduce recognition errors significantly and to provide a simple and yet similar image- The present invention is to provide a system for recognizing an augmented reality image using an overlap cut image and a method for recognizing the same.

According to an aspect of the present invention, there is provided an augmented reality image recognition system using an overlap cut image, comprising: a video device capable of inputting and outputting an image for a video image to be augmented; A reference image storage unit for sorting the feature regions of the image into overlapping image elements and constructing and storing the image regions by using an overlap-cut method; An image recognition unit for generating a histogram from image elements input from the image device and determining a similar image among the reference images stored in the reference image storage unit as a recognition image and enhancing the recognition image into an image or a 3D object model; Wherein the image recognition unit includes a primary recognition unit for performing a determination as a recognized image immediately when there is no similar or confused reference image in the recognition process for the image, When at least two images are recognized, the image element histogram of the image image is compared with the image element information of the reference images stored in the reference image storage unit, and the most similar reference image is searched to determine the recognition image 2 to perform identification It comprises recognition unit.

The reference image storing unit includes an element generating unit for generating an image element by selecting a unique feature portion of the corresponding reference image in the entire reference image region, an element extracting unit for extracting a feature point for each image element generated in the element generating unit, An element information storage unit for storing the information after generating the histogram and the feature points for each feature region extracted from each image element, including the identification number of the reference image, and an image information storage unit for storing, from the information of each image element stored in the element information storage unit, And an information inserting unit for displaying a geometric relative position of the corresponding element in the first area.

The image element configuration for the reference image in the element generation unit is made to be superimposable so as to generate a plurality of image elements with an emphasis on an area showing a difference from a similar image.

A histogram generator for generating a histogram of an image element of a reference image in the image image input from the image device; and a histogram generator for generating an element histogram of the image image generated from the histogram generator, An image determining unit that determines a reference image that includes the most similar image elements among the reference images retrieved from the image retrieving unit as a recognized image by comparing the element information of the image element with the element information of the image, And estimating a stereoscopic attitude of the recognized image selected from the reference image and calculating a homography of the image apparatus from the reference image on the reference image storage unit and the position of the virtual image Comprise the image estimator for estimating the position, the image enhancing unit for using the estimated posture information is completed by the image estimation image or enhancing the 3D object models to be represented on the picture of the video device.

The histogram generator may include a data generator for receiving a video image of the video apparatus and extracting a feature point of the image to generate a searchable data structure, And an image matching unit for comparing the feature points of the image elements with respect to the reference image of the storage unit.

The image determining unit may be configured to set a specific threshold value that is a determination criterion for recognition or not, determine only the reference image that satisfies the set threshold value as the recognized image, and determine the most similar image among the reference images that satisfy the set threshold value, .

The image recognition unit comprises an overlap-cut image recognition method for dividing an area of an image element of a reference image into variable-size and number of element areas.

According to another aspect of the present invention, there is provided a method of recognizing an augmented reality image using an overlap cut image, comprising: inputting a video image by capturing an image to be augmented; Constructing a histogram of image elements of a stored reference image by constructing the image to be an identification object in the image; Comparing the element histogram of the video image with the element information of the reference images to search for the most similar reference image; Determining a reference image including the most similar image elements among the reference images as a recognized image; Estimating the posture of the image selected as the recognized image in a stereoscopic manner, calculating homography of the camera from the video image and the reference image, and estimating the position and posture of the virtual camera; Augmenting the image or the 3D object model using the attitude estimation information of the recognized image; And outputting the enhanced image as an image capable of expressing the enhanced image from the recognized image.

The method comprising the steps of: constructing an image element selected for each region of a characteristic portion unique to the entire reference image by an overlap-cut method; extracting feature points for each image element of the reference image; Creating and storing feature points extracted from the image elements of each reference image, including the identification number and histogram of the reference images, and inserting the feature points in the reference image to display the geometric relative positions of the information of each image element.

In order to construct the image element for the reference image, a plurality of image elements are superimposed on an area showing a difference from the similarity of the corresponding image.

A first recognition step of performing a determination as a recognition image immediately when there is no similar reference image in the recognition process of the video image; and a step of recognizing at least two similar reference images in the recognition process of the video image And a secondary recognition step of comparing the histogram of the image element of the video image with the image element information of the reference images and then searching for the most similar reference image and performing identification to complete the determination as the recognized image.

Generating the histogram by extracting minutiae points of a video image and generating a searchable data structure; comparing the minutiae points of the video image having the data structure and minutiae points of the image elements with respect to the reference image to perform matching; .

When creating the histogram, the region for the image element of the reference image is divided into variable-size and number of element regions.

To determine the reference image as a recognized image, the most similar image is determined as a recognized image only in reference images that satisfy the set threshold value based on a specific set threshold value.

According to the augmented reality image recognition system and the recognition method using the overlap cut image according to the present invention, when there is no similar image, the information about the similar image is compared and retrieved together with the first recognition to complete the determination as the recognition image immediately Since the image recognition structure including the secondary recognition that completes the determination with the recognition image is constructed, the recognition accuracy can be improved by reducing the phenomenon of recognition from the similar image quickly while discriminating the image quickly.

In addition, the augmented reality image recognition system and the recognition method using the overlap cut image according to the present invention constitute image elements in a manner of overlapping each other with respect to regions showing similarities and differences in reference images, So that accurate image recognition can be achieved.

In addition, since the AR system and the recognition method using the overlap cut image according to the present invention are configured to set a specific threshold value as a decision criterion for image recognition, it is possible to block the recognition error due to a false recognition, It is possible to broaden the range.

Also, the ARP system and the recognition method using the overlap cut image according to the present invention configure an overlap-cut image recognition method for dividing the image region into a variable size and a number of regions, So that the false recognition rate can be minimized.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing an embodiment of an augmented reality image recognition system using an overlap cut image according to the present invention.
BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to a method and system for recognizing augmented reality image using an overlap cut image.
3 is a diagram schematically showing an element configuration of an element generation unit in an augmented reality image recognition system using an overlap cut image according to the present invention.
4 is a diagram schematically illustrating a histogram of an element information storage unit in an augmented reality image recognition system using an overlap cut image according to the present invention.
5 is a diagram illustrating a histogram of an image recognition unit in an augmented reality image recognition system using an overlap cut image according to the present invention.
6 is a diagram illustrating an image recognition method of an image recognition unit in an augmented reality image recognition system using an overlap cut image according to the present invention.
FIG. 7 is a block diagram illustrating an embodiment of a method of recognizing an augmented reality image using an overlap cut image according to the present invention.
FIG. 8 is a block diagram illustrating a process of constructing reference image information in an embodiment of a method of recognizing an augmented reality image using an overlap cut image according to the present invention.
FIG. 9 is a block diagram illustrating an image recognition process in an embodiment of an augmented reality image recognition method using an overlap cut image according to the present invention.

The present invention relates to an image enhancement apparatus, A reference image storage unit for sorting the feature regions of the image into overlapping image elements and constructing and storing the image regions by using an overlap-cut method; An image recognition unit for generating a histogram from image elements input from the image device and determining a similar image among the reference images stored in the reference image storage unit as a recognition image and enhancing the recognition image into an image or a 3D object model; Wherein the image recognition unit includes a primary recognition unit for performing a determination as a recognized image immediately when there is no similar or confused reference image in the recognition process for the image, When at least two images are recognized, the image element histogram of the image image is compared with the image element information of the reference images stored in the reference image storage unit, and the most similar reference image is searched to determine the recognition image 2 to perform identification The augmented reality image recognition system using the overlapped-cut image including recognition unit, characterized in technology configuration.

According to another aspect of the present invention, Constructing a histogram of image elements of a stored reference image by constructing the image to be an identification object in the image; Comparing the element histogram of the video image with the element information of the reference images to search for the most similar reference image; Determining a reference image including the most similar image elements among the reference images as a recognized image; Estimating the posture of the image selected as the recognized image in a stereoscopic manner, calculating homography of the camera from the video image and the reference image, and estimating the position and posture of the virtual camera; Augmenting the image or the 3D object model using the attitude estimation information of the recognized image; And outputting the enhanced image as an image capable of expressing the enhanced image from the recognized image.

Next, a preferred embodiment of an augmented reality image recognition system using an overlap cut image according to the present invention will be described in detail with reference to the drawings.

1 and 2, an embodiment of an augmented reality image recognition system using an overlap cut image according to the present invention includes an image device 10, a reference image storage unit 20, an image recognition unit 30).

The video equipment 10 is provided with an image input / output capability for a video image to be augmented. That is, the video apparatus includes a device capable of capturing an image of a print book and inputting the captured image, and a monitor capable of outputting an image including the captured image on the screen.

The printed book includes all the offline books including at least one page including text or pictures, and the page of the printed book has at least one marker for augmented reality.

The video device 10 includes a camera and a monitor screen such as a portable multimedia player (PMP), a personal digital assistant (PDA), a mobile phone, a smart phone, etc., Is a concept collectively referred to as a device having a predetermined storage function and an arithmetic function.

The reference image storage unit 20 constitutes an image element for each of a plurality of image images to be recognized by the image device 10, and then constructs the reference image as a reference image to be identified.

As shown in FIG. 2, the reference image storing unit 20 is configured to be able to store classified image elements classified into individual reference images and store the reference image. The reference image storing unit 20 includes an element generating unit 21, an element extracting unit 23, An element information storage section 25, and an information inserting section 27.

The element generation unit 21 forms an individual divided reference image by selecting an area that can be characterized for each image with respect to all reference images to be identified. That is, the element generating unit 21 selects a characteristic feature portion of the corresponding reference image among the entire reference images, and constructs a reference image classified into the image element.

In this case, one image element is a part of the entire image of the reference image, which is a part of the image so as to display the feature of each corresponding image. Therefore, the part including the unique feature of the image is selected to constitute the image element.

In the reference image storage unit 20, the element generating unit 21 constructs an image element of a reference image, and the element generating unit 21 selects an image feature region in an overlap-cut manner with respect to a reference image.

As shown in FIG. 3, the image element configuration for the reference image in the element generation unit 21 is an overlap-cut method, and can be superposed so as to generate a plurality of image elements with an emphasis on an area showing a difference from a similar image. In other words, a plurality of image elements are generated which mainly overlap an area showing a difference from a similar image of the reference image.

The method of constructing the image element is selected in a heuristic manner so that it can be distinguished and stored according to the characteristics of each image according to the contents of the reference image.

The element generating unit 21 may be configured with one or two simple image element regions that are not overlapped if there is no reference image that is similar or confused with each other when the image element for the reference image is created.

The element extracting unit 23 extracts feature points for each image element generated by the element generating unit 21.

As an extraction method of the element extracting unit 23, an algorithm of a method most suitable to the characteristics of an application may be selected among a variety of algorithms such as a scale invariant feature transform (SIFT), a speeded up robust features (SURF) do.

The element information storage unit 25 stores the information after generating the histogram and the feature points for each feature region, including the identification number of the reference image, from each image element.

For example, as shown in FIG. 4, the element information storage unit 25 stores an identification number divided for each reference image and stores the information about the minutiae of the image elements for each reference image, and generates a histogram do.

The information inserting unit 27 displays the geometric relative position of the corresponding element in the reference image from the information of each image element stored in the element information storing unit 25. [

If the geometric relative position is displayed as described above, it can be used as additional information for further increasing the recognition rate according to the image recognition process, and insertion of such information can minimize the false recognition rate.

The image recognition unit 30 enhances the image input from the image device 10 into an image or a 3D object model and transmits the image in a form that can be output on a monitor screen of the image device 10. [

The image recognition unit 30 is configured to recognize a similar image among reference images based on a video image input from the video device 10 as a recognition image. That is, the image recognition unit 30 generates a histogram from image elements in the image image, and then determines a similar image among the reference images stored in the reference image storage unit 20 as a recognition image, and enhances the recognition image into an image or a 3D object model .

1, the image recognition unit 30 divides the recognition process of a video image input from the video equipment 10 into a primary recognition unit 30a and a secondary recognition unit 30b, So that the image can be recognized.

The primary recognition unit 30a performs image recognition so as to complete the determination as a recognized image immediately if there is no similar or confused reference image in the process of recognizing the image. For example, when the primary image recognition unit 30a compares a reference image stored on the reference image storage unit 20 with a video image input from the video apparatus 10, When only one reference image is searched for, it is determined to be a recognized image immediately.

By constructing the primary recognition unit 30a as described above, it is possible to complete a quick identification operation on the recognized image.

The secondary recognition unit 30b is provided when at least two similar reference images are recognized in the process of recognizing a video image.

The secondary recognition unit 30b compares the image element histogram of the video image with the image element information of the reference images stored in the reference image storage unit 20 and then searches for the most similar reference image to determine a recognition image Lt; / RTI >

1 and 2, the image recognition unit 30 receives a video image from the video device 10 and includes a structure capable of interlocking with the reference image storage unit 20 so that a reference image can be retrieved And includes a histogram creating unit 31, an image searching unit 33, an image determining unit 35, an image estimating unit 37, and a video enhancing unit 39.

The histogram creating unit 31 is configured to create a histogram of the image elements of the reference image in the image image input from the imaging device 10. [

The histogram generating unit 31 includes a data generating unit 31a that receives a video image of the video apparatus 10 and extracts minutiae points of the image to generate a searchable data structure. That is, the data generation unit 31a generates a descriptor by extracting a minutiae from a frame unit image received from the video equipment 10, thereby generating a search data structure.

It is preferable that the data generator 31a is configured to use a data structure such as a K-D tree (k-dimensional tree) or a FLANN tree.

The histogram generating unit 31 is provided with an image matching unit 31 for comparing the minutiae points of the frame image having the data structure completed by the data generating unit 31a with the minutiae points of the image elements for the reference image of the reference image storing unit 20. [ (31b).

As shown in FIG. 5, the image matching unit 31b creates a histogram of the image element of the reference image matched in the input image.

Each image element in the image recognition unit 30 functions as one element for recognizing a video image input from the video apparatus 10. [

The image searching unit 33 is configured to search the reference image storage unit 20 for a reference image most similar to a video image input from the video apparatus 10. That is, the image retrieval unit 33 retrieves the most similar reference image by comparing the element histogram of the image image created by the histogram creation unit 31 with the element information of the reference images stored in the reference image storage unit 20.

The image determining unit 35 determines a recognized image among the reference images retrieved from the image retrieving unit 33, and determines a reference image including the most similar image elements among the reference images as a recognized image.

The image determining unit 35 is configured to set a specific threshold value, which is a criterion for determining whether or not the image is recognized.

The image judging unit 35 judges only the reference image which satisfies the set threshold value as the recognized image.

For example, only the reference image that is the reference image that is searched with the most similar image among the reference images and that satisfies the threshold value set on the image determination unit 35 is determined as the recognized image. That is, it is determined that the reference image having the highest similarity among the reference images satisfying the threshold value set in the image determining section 35 is a recognized image, and if there is no reference image that satisfies the set threshold value, it is determined that the reference image is unrecognizable.

If the degree of similarity is higher than that of the other reference images, it is possible to prevent the phenomenon that the reference image is unconditionally recognized as the corresponding image even though the reference image is different from the image, thereby reducing the false recognition rate It is possible.

The image estimating unit 37 and the image enhancing unit 39 are configured to perform a tracking process after the image recognition is completed through the image determining unit 35. At this time, if the tracking fails, the histogram creating unit 31 again performs image matching to attempt image recognition.

The image estimating unit 37 estimates the attitude of the recognized image selected from the image determining unit 35 in a stereoscopic manner.

The image estimating unit 37 calculates the homography of the video apparatus 10 from the video image input from the video apparatus 10 and the reference image on the reference image storing unit 20, To estimate the position and posture.

The image enhancer 39 enhances the image or the 3D object model to be displayed on the image of the imaging device 10 using the completed posture estimation information in the image estimator 37. [

6, in the image recognition unit 30, the image recognition unit 30 divides an area for an image element of a reference image into variable-size and number-of-element regions, .

For example, when the image element for the reference image is created, the image recognition unit 30 may include one or two simple image element regions that are not overlapped if there is no similar or confused reference image, The expected reference image constitutes an image element with an overlap-cut recognition method in which an area showing a difference from a similar image is mainly emphasized.

As described above, when the overlapping element region is divided and superimposed on the region showing the difference from the similar image using the overlap-cut recognition method, the number of elements showing the unique characteristic of the image is increased. Therefore, It is possible to enhance the probability of being separately recognized.

Next, a preferred embodiment of a method for recognizing an augmented reality image using an overlap cut image according to the present invention will be described in detail with reference to the drawings.

As shown in FIG. 7, an embodiment of a method of recognizing an augmented reality image using an overlap cut image according to the present invention includes a step (S10) of inputting a video image, a step (S20) of creating a histogram, A step S40 of determining a recognized image, a step of estimating a posture S50, a step of enhancing the pitch S60, and a step S70 of outputting a video image .

In order to recognize an image, a reference image, which is an identification target, is constructed to be input together with a video image to be augmented.

In order to construct the information on the reference image, first, a unique feature portion of the corresponding reference image is selected for each region to constitute a reference image classified into the image elements (S1)

In order to construct the image element for the reference image, an overlap-cut method is used to generate a plurality of image elements in an overlapping manner to generate a plurality of image elements with an emphasis on an area showing a difference from a similar image.

After constructing the image elements of the reference image, feature points for each image element of the reference image are extracted (S3)

In extracting the image elements of the reference image, algorithms of various methods such as SIFT, SURF, and FAST are used.

After extracting the feature points of the image elements of the reference image, feature points extracted from the image elements of each reference image, including the identification number and the histogram of the reference image, are generated and stored (S5)

In order to construct the information about the reference image, a geometric relative position of the information of each image element is inserted in the reference image. (S7)

In the step of inputting the image image (S10), the image of the object to be augmented, that is, the image of the printed book, is imaged by the image device and the image is inputted.

In the above, a small-sized device such as a PMP, a PDA, a mobile phone, or a smart phone is used as a video device.

In the step of creating the histogram (S20), a histogram of the image elements of the stored reference image is constructed by constructing the histogram as an identification object in the image.

In order to create the histogram, the feature points of the image transferred from the imaging device are extracted and generated as a searchable data structure (S21).

It is preferable that the data structure of the video image is generated by a data structure such as a K-D tree or a FLANN tree.

In the process of creating the histogram, the feature points of the image image having the data structure are compared with the feature points of the image elements of the reference image to perform matching (S23). That is, to generate the histogram, Create an image element of the image as a histogram.

When creating the histogram, the region for the image element of the reference image is divided into variable-size and number of element regions.

For example, if you do not have similar or confusing reference images when you create an image element for a reference image, you can divide it into one or two simple image element regions that do not overlap, and the reference image, So that the region showing the difference is superimposed on the central point.

In step S30 of retrieving the reference image, the most similar reference image is searched by comparing the element histogram of the image with the element information of the reference images.

In the determination step S40, the reference image including the most similar image elements among the reference images is determined as the recognized image.

In the above, the determination is made based on the specific set threshold value. That is, the most similar image is determined as the recognized image only in the reference image that satisfies the set threshold value based on the specific set threshold value.

For example, only the reference image that is the reference image that is searched for the most similar image among the reference images and satisfies the set threshold value is determined as the recognized image. That is, it is determined that the reference image that has the highest similarity among the reference images satisfying the set threshold value is determined as the recognized image, and if there is no reference image that satisfies the set threshold value, it is determined that the reference image is unrecognizable.

In addition, the recognition process for the image is performed twice. In other words, if there is no similar reference image in the recognition process of the video image, a first recognition step (S45) is performed to complete the determination as a recognition image immediately. In the recognition step of the video image, A second recognition step S47 for comparing the histogram of the image element of the video image with the image element information of the reference images and then searching for the most similar reference image and performing the identification to complete the determination as the recognition image.

The step of estimating the posture (S50) and the step of enhancing (S60) are performed to proceed with the tracking process when the image recognition is completed in the step of determining as the recognized image (S40).

At this time, if the tracking fails, the image is matched again through the step of creating the histogram (S20) to retry the image recognition.

In the step of estimating the posture (S50), the posture of the image selected as the recognized image is three-dimensionally estimated. In other words, to estimate the posture, the homography of the camera is calculated from the image and the reference image, and the position and posture of the virtual camera are estimated.

In the enhancing step S60, the image or the 3D object model is augmented using the attitude estimation information of the recognized image.

In step S70, the enhanced image is transmitted to the image device and the image is output to the user.

That is, according to the augmented reality image recognition system and the recognition method using the overlap cut image according to the present invention configured as described above, when there is no similar image, It is possible to improve the recognition accuracy by reducing the phenomenon of false recognition from the similar image while quickly identifying the image, since the image recognition structure having the secondary recognition that completes the determination as the recognition image after comparing and searching the information is constructed.

In addition, since the present invention forms image elements in a superimposed manner with emphasis on the regions showing similarities and differences in the reference image, accurate image recognition can be achieved by increasing the frequency of image recognition.

Further, the present invention is configured to set a specific threshold value, which is a determination criterion for image recognition, so that it is possible to block the recognition error due to false recognition and extend the application range of the existing augmented reality to a wide range.

Although the preferred embodiments of the augmented reality image recognition system and the recognition method using the overlap cut image according to the present invention have been described above, the present invention is not limited thereto. It is possible to carry out the modification, and this is also within the scope of the present invention.

10: video equipment 20: reference image storage unit
21: element generating section 23: element extracting section
25: element information storage unit 27: information insertion unit
30: image recognition unit 30a: primary recognition unit
30b: secondary recognition unit 31: histogram preparation unit
31a: data generating unit 31b: image matching unit
33: image retrieving unit 35: image judging unit
37: image estimating unit 39: image enhancing unit

Claims (14)

A video device provided for inputting and outputting an image of a video image to be augmented;
A reference image storage unit for sorting the feature regions of the image into overlapping image elements and constructing and storing the image regions by using an overlap-cut method;
An image recognition unit for generating a histogram from image elements input from the image device and determining a similar image among the reference images stored in the reference image storage unit as a recognition image and enhancing the recognition image into an image or a 3D object model; Including,
Wherein the image recognizing unit includes a primary recognizing unit for performing a determination as a recognized image immediately when there is no similar or confusing reference image in the process of recognizing a video image, Comparing the image element histogram of the image image with the image element information of the reference images stored in the reference image storing unit, searching for the most similar reference image, and identifying And a second recognition unit for performing the second recognition,
The reference image storing unit may include an element generating unit that generates an image element by selecting a unique feature portion of the corresponding image among the entire reference images by region and an element extracting unit that extracts a feature point for each image element generated by the element generating unit An element information storage unit for storing information after generating a histogram and feature points for each feature region extracted from each image element including an identification number of a reference image, And an information insertion unit for displaying a geometric relative position of the corresponding element in the image.
delete The method according to claim 1,
Wherein the image element configuration for the reference image in the element generation unit is made to be superimposable so as to generate a plurality of image elements with an emphasis on an area showing a difference from a similar image.
The method according to claim 1,
The image recognizer may include a histogram generator for generating a histogram of image elements of a reference image in a video image input from the video apparatus, a histogram generator for generating an element histogram of the video image generated by the histogram generator, An image retrieving unit for retrieving the most similar reference image by comparing with the element information of the images; an image judging unit for judging, as a recognized image, a reference image including the most similar image elements among the reference images retrieved by the image retrieving unit; And estimating stereoscopically the attitude of the recognized image selected from the determination unit, calculating homography of the video apparatus from the video image input from the video apparatus and the reference image on the reference image storage unit, And an image enhancement unit for enhancing the image or the 3D object model to be displayed on the image of the image device using the posture estimation information completed in the image estimating unit, using an overlap cut image Realistic image recognition system.
The method of claim 4,
The histogram creating unit includes a data generating unit for receiving a video image of the video apparatus and extracting a minutiae point of the image to generate a searchable data structure, And an image matching unit for comparing the feature points of the image elements with respect to the reference image of the image storing unit.
The method of claim 4,
The image determining unit may be configured to set a specific threshold value that is a determination criterion for recognition or not, determine only the reference image that satisfies the set threshold value as the recognized image, and determine the most similar image among the reference images that satisfy the set threshold value, Of the overlap-cut image.
The method according to claim 1,
Wherein the image recognition unit comprises an overlap-cut image recognition method for dividing an area of an image element of a reference image into variable-size and number-of-element regions.
A step of capturing an image to be augmented and inputting a video image;
Constructing a histogram of image elements of a stored reference image by constructing the image to be an identification object in the image;
Comparing the element histogram of the video image with the element information of the reference images to search for the most similar reference image;
Determining a reference image including the most similar image elements among the reference images as a recognized image;
Estimating the posture of the image selected as the recognized image in a stereoscopic manner, calculating homography of the camera from the video image and the reference image, and estimating the position and posture of the virtual camera;
Augmenting the image or the 3D object model using the attitude estimation information of the recognized image;
And outputting the enhanced image as an image capable of expressing the enhanced image from the recognized image,
In constructing a histogram for an image element of a stored reference image by constructing and storing an image of the reference image in the image, constructing and storing the image element of the reference image may include storing the entire reference image in an overlap- A step of extracting feature points for each image element of the reference image, a step of generating and storing the feature points extracted from the image elements of each reference image including the identification number and the histogram of the reference image, And inserting the overlapping cut image to display a geometric relative position of the information of each image element in the reference image.
delete The method of claim 8,
In the step of constructing the image elements selected by the overlapping cut method for each of the regions of the feature portion, the plurality of image elements are superimposed on the region showing the difference from the similarities of the images, And an overlap-cut image is generated.
The method of claim 8,
Wherein the step of determining a reference image including the most similar image elements among the reference images as a recognition image includes a first recognition step of performing a determination as a recognition image immediately when there is no reference image similar to each other, If at least two similar reference images are recognized in the process, the histogram of the image element of the image image is compared with the image element information of the reference images, and the most similar reference image is searched to perform the identification to complete the determination as the recognition image A method for recognizing an augmented reality image using an overlap cut image including a second recognition step
The method of claim 8,
The step of generating the histogram comprises the steps of extracting minutiae points of a video image to generate a searchable data structure, comparing the minutiae points of the video image having the data structure with the minutiae points of the image elements of the reference image, The method of claim 1, further comprising:
The method of claim 8,
Wherein an area for the image element of the reference image is divided into a variable size and a number of element areas when the histogram is created.
The method of claim 8,
Wherein the determination of the reference image as a recognition image is made by determining an image that is most similar to only a reference image satisfying a set threshold value based on a specific set threshold value as a recognition image.

Images