KR20070080066A - System for personal authentication and electronic signature using image recognition and method thereof - Google Patents

Abstract

A system and a method for authenticating a user and an electronic signature with image recognition are provided to realize a safe and stable authentication system in a mobile terminal by using a portable image medium and a transmitter. A mobile terminal(10) captures and transmits the image used as an authentication image by user operation. An image recognition server(20) estimates projection transformation relation between an image stored in a database and the image received from the mobile terminal by extracting/comparing SIFT(Scale Invariant Feature Transform) feature vector from the received image with the SIFT feature vector of the authentication images stored in the database, and using position relation of the SIFT feature vectors. An authentication server(30) transmits a result to the mobile terminal by determining the unique image of the user and authenticating the user/electronic signature based on the estimated projection transformation relation.

Description

System for Personal Authentication and Electronic Signature Using Image Recognition And Method Thereof}

1 is a block diagram showing the concept of a personal authentication and digital signature system using image recognition according to the present invention,

2 is an operation flowchart for explaining a personal authentication and digital signature method using image recognition according to the present invention;

3 is a diagram illustrating a method of constructing a DOG image pyramid to extract a SIFT feature vector;

4 is a conceptual diagram illustrating 26 adjacent pixel positions compared to extract key positions, and

FIG. 5 is a conceptual diagram of extracting a directional distribution for an 8x8 block and quantizing it in eight directions.

<Description of Signs for Main Parts of Drawings>

10: mobile terminal 20: video recognition server

30: authentication server

The present invention relates to a personal authentication and digital signature system using image recognition and a method thereof, and more particularly, to a personal authentication and electronic signature system using image recognition to perform personal signature and authentication using a unique image of the individual and It's about how.

Due to the rapid increase in security systems, card use, and electronic commerce, secure and accurate authentication system through communication media has become a very important technical element.

Generally, a few digit password or personal signature is used to confirm and approve an individual.

However, this method has a problem that it can be easily exposed to the outside during transmission or use, and is easy to steal because the data structure is simple. In the case of performing authentication through biometrics, it is possible to recognize the biometric information to be recognized so as to be as similar as possible in the database. In order to perform face recognition, it is necessary to acquire images with as few external shadows or lighting as possible in front of the faceless impression.

Therefore, in a general real data acquisition environment, in order to perform recognition and authentication using a biometric system, accurate use must be induced to the user, and the actual recognition result is often much lower than the experimental result.

In addition, when recognizing an object having a specific shape or pattern, it is necessary to first divide exactly which part of the object is the object in the image, and such image segmentation is also a very difficult problem and is not practical.

Recently, a system for performing authentication using 2D barcodes has been developed. In the case of 2D barcodes, many problems of image recognition have been solved. If the resolution is not maintained enough, there is a problem that the pattern of the barcode is not recognized.

In addition, passwords and text signatures can be easily exposed to outside or stolen during the transmission process, there is a problem that the recognition performance is greatly degraded when the bio-signals are not obtained in an accurate and normalized form.

Accordingly, the present invention has been made to overcome the above problems, and unlike the conventional authentication system that transmits and recognizes passwords, text signatures, and biological signals through a communication medium, it provides a stable authentication system based on image recognition. Its purpose is to.

In addition, the present invention solves various problems of the existing authentication system and provides a more secure and stable authentication system by using a portable video medium and a transmission device that is currently widely used, and applying a technique for recognizing a general image instead of a specific pattern. Its purpose is to.

In order to prevent the recognition rate from being degraded due to the difficulty of image pattern and object-based image recognition in the prior art and the inaccuracy of the preprocessing process, the present invention recognizes the image as a whole rather than the object based on the specific region of the image. Then, it uses characteristic elements that can be recognized regardless of changes in image size, projection transformation, lighting, and the like.

In addition, the present invention does not require a complicated and inaccurate preprocessing process such as object segmentation, and uses SIFT, which is stably extracted and recognized for image resolution, size, illumination, and projection distortion, as a basic feature of image recognition.

In order to recognize the same image, basically, how similar and how many SIFT feature vectors are extracted from the database image and the acquired image are considered. Then, by extracting the geometric transformation, that is, the projection transformation matrix (homography) between the two images using the positions of the corresponding SIFT feature vectors, it is verified that the image is the same image even though the geometric transformation has been performed. In order to derive an optimal projection transformation matrix, least squares are used using corresponding positional relations.

In order to achieve the above object, the present invention provides the characteristic technical elements as follows.

First, full recognition of general image, not specific pattern or object, second, image retrieval using SIFT feature vector, third, image recognition considering geometric distortion through projection transformation (homography) estimation, fourth, 128-dimensional real number coefficient A vector quantization technique that transforms SIFT feature vectors into 16-dimensional integer coefficient vectors.

According to a first aspect of the present invention for achieving the above object The personal authentication and digital signature system using image recognition includes : A) a mobile terminal which photographs and transmits an image to be used as an authentication image by a user's operation, and B) extracts a SIFT feature vector from the transmitted image, And an image recognition server for estimating the projection transformation relationship between the image of the database and the obtained image by using the positional relationship of the SIFT feature vectors while simultaneously comparing the SIFT feature vectors of the authentication images stored in the database. And an authentication server for determining a unique image of the individual by using the projection conversion relationship, performing personal authentication and signature based on this, and transmitting the result to the mobile terminal.

Here, preferably, the authentication image may be a Photoshop image, etc., produced by the user himself, not the photographed image.

In addition, the personal authentication and digital signature method using the image recognition according to the second aspect of the present invention for achieving the above object, a) using the mobile terminal to obtain the authentication image and at the same time transmitting the obtained authentication image Step b) recognizing the transmitted image using the image previously stored in the image recognition server; and c) the authentication server determines the unique image of the individual based on the recognized image and performs personal authentication and signature accordingly. And then transmitting the result to the mobile terminal.

Considering the problem of the existing authentication system, the present invention requires an image recognition based recognition system having the following characteristic differences. First, it does not require complex and unstable image processing techniques such as segmentation or edge detection. Second, the image must be stably recognized regardless of various changes in the image such as rotation, lighting, geometric distortion, resolution, and size change. Third, it must be possible to acquire data images without requiring a high degree of sophistication or skill.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram illustrating the concept of a personal authentication and digital signature system using image recognition according to the present invention.

As shown in the drawing, the system acquires an SIFT feature vector from the mobile terminal 10 and the transmitted image at the same time a user who wants to perform authentication acquires an authentication image using the mobile terminal 10. The image recognition server extracts and compares it with SIFT feature vectors of authentication images pre-stored in a predetermined database and estimates the projection transformation relationship between the image of the database and the acquired image using the positional relationship of the SIFT feature vectors. And an authentication server 30 for determining a unique image of the individual by using the estimated projection transformation relationship, performing personal authentication and signature based on this, and transmitting the result to the mobile terminal 10. do. In this case, the authentication image may be a Photoshop image produced by the user himself, not a photographed image.

The mobile terminal 10, an image input unit (not shown), a transmission device (not shown) for receiving the captured image data and transmitting it to the server 20 through the wireless module, the image transmission process and And a display unit (not shown) for receiving product information and displaying the same on a screen.

The input unit includes at least one button input means (not shown).

Hereinafter, the operation principle of the personal authentication and digital signature system using image recognition according to the present invention configured as described above will be described with reference to FIGS.

2 is a flowchart illustrating a personal authentication and digital signature method using image recognition according to the present invention.

First, a user photographs and acquires an image to be used as an authentication image by using the mobile terminal 10 (S31). In this case, the mobile terminal 10 downloads a program (for example, an authentication image transmission program, etc.) necessary for an operation of acquiring and transmitting an authentication image on its terminal screen by accessing a specific server of a mobile communication provider through a mobile communication network. At this time, the user can download while watching through the terminal screen. In addition, the mobile terminal 10 may charge the cyber money for providing authentication information according to the authentication image through the mobile communication network, or may perform the cyber money charging after the authentication process through the wired network of the PC. However, the above description is well known to those skilled in the art and will be readily appreciated.

Thereafter, the image recognition server 20 receives an image obtained from the mobile terminal (S32), extracts an SIFT feature vector from the transmitted image, and then extracts the SIFT feature vector of authentication images stored in a predetermined database. Each comparison is made (S33). In this case, the similarity between vectors is basically defined by the distance of the vector space, and it is determined that the SIFT feature vectors that are very close to each other at a certain level correspond to each other.

Thereafter, the image recognition server 20 estimates the projection transformation relationship between the pre-stored image of the database and the acquired and transmitted image by using the positional relationship of the corresponding SIFT feature vectors (S34). In this case, a database image having a large number of corresponding SIFT feature vectors and having high matching between two images by an estimated projection transformation is determined as a unique image of a corresponding individual.

Thereafter, the authentication server 30 performs personal authentication and signature through the determined unique image and transmits the result to the mobile terminal 10 (S35). In this case, the user may store his or her own video in a mobile terminal or own it in the form of a picture card and use it for authentication along with a password.

Hereinafter, a technique of extracting the aforementioned SIFT feature vector applied to the present invention will be described.

First, the original image is passed through a Gaussian filter having variance values of various scales, and the resulting images are sequentially subtracted from each other according to adjacent scales. This difference image is called a Gaussian difference (DOG) image. The DOG image is subsampled 2: 1 and passed through a Gaussian filter with various scales. As shown in FIG. 3, after obtaining DOG images of a pyramid structure through this process, a core position corresponding to a maximum or minimum of each DOG image is found.

As shown in FIG. 4, the maximum or minimum position is selected as the maximum or the minimum in comparison with 26 pixels, in which each of the surrounding 8 pixels and the 9 pixels of the up and down scale DOG image are all. In order to remove the noise-induced elements, key positions on strong edge lines or small DOG values are removed. A 16x16 block is set around the last selected key position, and the directionality is measured for the pixels in the block. The direction and the size of each pixel are measured to estimate the direction that is largest in the block. Set this direction to the direction of its core position. In order to obtain the final SIFT feature vector, 16x16 blocks are set again around each core position, and then divided into 4x4 block units to measure the direction of change of pixel values for each pixel in the block. The change direction is set to be proportional to the change slope to obtain a probability distribution with respect to the direction so that the larger the magnitude of the change, the greater the probability distribution of the corresponding direction.

As shown in FIG. 5, the directional probability distribution is stored in eight directions, which are represented in a 128-dimensional vector form for a 16 × 16 block. The vector representing the directional probability distribution is a structure of the SIFT feature vector, and is stored together with parameters such as the image coordinate of the core position, the directionality of the center position, and the Gaussian scale. The 128-dimensional SIFT feature vector thus obtained is represented by a codebook by performing vector quantization.

The initially obtained SIFT feature vector is an arbitrary real number with a floating point, which requires a relatively large transmission capacity and time when transmitting the SIFT feature vector or comparing distances. Therefore, in order to overcome this disadvantage, vector quantization is performed on 8-dimensional vectors of SIFT divided into 8-direction units, and 8-dimensional direction distributions represented by real values are mapped to 1-dimensional codebooks. Through this vector quantization process, a 128-dimensional real coefficient SIFT feature vector is converted into a 16-dimensional integer coefficient vector. The 16-dimensional vector quantized SIFT feature vector can be transmitted with a lower data amount. When calculating the difference between the vectors, the 16-dimensional vector quantized SIFT feature vector is reduced and compared to the 128-dimensional vector of the quantized codebook.

Estimating image distortion by projection transformation is an important way to improve image recognition performance. In addition to determining the number of local SIFT feature vectors of two images, it is necessary to determine how consistently each feature vector is arranged.

For example, suppose an image has m points matched for one image and n points matched for another image. If the m points of the former have no regularity, but the n points of the latter can be accurately explained by an Affine transformation or homography, it is reasonable to assume that the latter is an exact match, even if m is greater than n. In particular, if the size of the image is different and the shooting environment is different, the number of feature points extracted from it will change a lot, so it is not enough to simply count the number of matched points through distance comparison.

Therefore, it is necessary to count the number of points to be matched precisely from the viewpoint of transformation at the same time as the method described above. In the present invention, a task of estimating a change between the two images from the points obtained through the search of the nearest neighbor finding (Nearest Neighbor Finding) is added. The transform estimation used RANSAC (Random Sample Consensus) algorithm. That is, four points (in the case of homography) are randomly selected between the corresponding points, and the four points are used to predict the homography generated as follows.

Homography is a transformation that represents the relationship between multiple images when a plane is taken with cameras in different directions and at different distances, so the camera is suitable for the purpose of retrieving the original image by taking a video or picture played on the monitor. to be. By homography, the (X, Y) point before the first transformation is

Will be moved to.

Based on the transform estimated above, the pair of points that can be described by this transformation is called 'inlier' and the pair of points that are not is called 'outlier'. In each iteration, the number of 'inliers' is counted. After repeating this process several times, the largest number of 'inliers' is the result of the best prediction of the transformation between the two images, where the number of 'inliers' indicates the match between the two images we found. Exponent. Doing this for all images belonging to the search target is time-consuming, so do this only for images that show a large number of matches above a certain level in the results obtained through the shortest distance search. Perform.

Such an authentication system can be applied to any organization that performs personal authentication, such as electronic commerce, card transactions, security system access, and the like. If image processing can be performed at the step of acquiring an image, authentication may be performed by transmitting the image or transmitting only the extracted SIFT feature using vector quantization.

Although the present invention has been described as a specific preferred embodiment, the present invention is not limited to the above-described embodiments, and the present invention is not limited to the above-described embodiments without departing from the gist of the present invention as claimed in the claims. Anyone with a variety of variations will be possible.

For example, according to the present invention, whenever the user wants to be authenticated, the image stored in the mobile terminal is transmitted to the authentication system (that is, the image recognition server 20 and the authentication server 30) to perform a comparison, and the previously stored image. If the service is performed by wireless like M bank, it can be authenticated after transmitting the authentication image owned by the user to the authentication system by using the wireless Internet. Alternatively, in the case of being authenticated offline, the company may perform authentication by directly photographing the image presented by the user and transmitting the image to the authentication system.

Although described above with reference to a preferred embodiment of the present invention, those skilled in the art will be variously modified and changed within the scope of the invention without departing from the spirit and scope of the invention described in the claims below. I can understand that you can.

As described above, according to the present invention, it is expected to effectively reduce problems such as error and number exposure that occur in the process of entering or signing a password for authentication. In other words, by uniquely using personalized unique images, it is less likely to be easily exposed or stolen by others, and because general images are used instead of regular patterns such as numbers and letters, the transmission process and the external exposure from the server can be reduced. I think there will be.

In addition, unlike the existing 2D barcode authentication method, even if a part of the original image is lost or the resolution is low, there is no problem in performing the recognition. It can be used as a very convenient and friendly system for users.

In addition, the present invention can realize a more secure and personalized personal authentication and electronic signature system based on a portable video system, and has a structure for acquiring and transmitting images using a mobile phone, such as mobile phone cameras and video related industries It is expected to play a role in increasing demand.

Claims (19)

A) a mobile terminal for photographing and transmitting an image to be used as an authentication image by a user's operation; B) Extract the SIFT feature vector from the transmitted image and compare it with the SIFT feature vector of the authentication images pre-stored in a predetermined database and at the same time use the positional relationship of the SIFT feature vectors to obtain the image and the acquired image. An image recognition server estimating a projection transformation relationship between images; And C) using an image recognition, characterized in that it comprises an authentication server for determining the unique image of the individual using the estimated projection conversion relationship, and based on this personal authentication and signature, and transmits the result to the mobile terminal Personal authentication and digital signature system. The method of claim 1, wherein the authentication video, Personal authentication and digital signature system using image recognition, characterized in that the image itself is a Photoshop image produced by the user himself. The method of claim 1, wherein the mobile terminal, And an image input unit, a transmission device that receives the captured authentication image data and transmits the received authentication image data to the image recognition server through a wireless module, and a display unit which receives the face image transmission process and information corresponding thereto and displays it on the screen. Personal authentication and digital signature system using recognition. The method of claim 3, wherein the image input unit, Personal authentication and electronic signature system using image recognition using image search, characterized in that it comprises at least one button input means. a) a mobile terminal for photographing and transmitting an image to be used as an authentication image by a user's operation; b) an image recognition server for estimating a projection transformation relationship between an image previously stored in its database and the transmitted image; And c) an authentication server for determining a unique image of the individual using the estimated projection transformation relationship, performing a personal authentication and a signature based on this, and transmitting the result to the mobile terminal. Personal authentication and digital signature system. The method of claim 5, wherein the mobile terminal, And an image input unit, a transmission device that receives the captured authentication image data and transmits the received authentication image data to the image recognition server through a wireless module, and a display unit that receives the authentication image transmission process and information corresponding thereto and displays it on the screen. Personal authentication and digital signature system using image recognition. The method of claim 6, wherein the image input unit, Personal authentication and digital signature system using image recognition, characterized in that it comprises at least one button input means. The method of claim 5 or 6, wherein the mobile terminal, Personal authentication and digital signature system using image recognition, characterized in that to download the program necessary for the operation of the authentication image acquisition and transmission on the terminal screen of the mobile communication network connected to a specific server of the mobile communication provider. A) acquiring an authentication image using a mobile terminal and simultaneously transmitting the obtained authentication image; B) recognizing the transmitted image using the image previously stored in the image recognition server; And C) the authentication server determines a unique image of the individual based on the recognized image, performs personal authentication and signature accordingly, and transmits the result to the mobile terminal. Authentication and digital signature method. The method of claim 9, wherein the authentication video, Personal authentication and digital signature method using the image recognition, characterized in that the image is a Photoshop image produced by the user himself rather than the photographed image. The method of claim 9, wherein the mobile terminal, Personal authentication and digital signature method using image recognition, characterized in that to download the program necessary for the operation of the authentication image acquisition and transmission on the terminal screen of the mobile communication network connected to a specific server of the mobile communication provider. The method of claim 9 or 11, wherein the mobile terminal, Personal authentication and digital signature using video recognition, which can charge cyber money for providing information according to authentication video through mobile communication network, or perform cyber money charging after authentication process through wired network of PC Way. 1) photographing and obtaining an authentication image through a mobile terminal; 2) receiving, by an image recognition server, an image obtained from a mobile terminal, extracting an SIFT feature vector from the transmitted image, and comparing the image with SIFT feature vectors of authentication images previously stored in a predetermined database; 3) The image recognition server estimates the projection transformation relationship between the pre-stored image of the database and the acquired and transmitted image by using the positional relationship of the corresponding SIFT feature vectors, and the unique image of the individual by the estimated projection transformation. Judging by; And And 4) performing, by the authentication server, personal authentication and signature through the determined unique image, and transmitting the result to the mobile terminal. The method of claim 13, wherein the authentication video, Personal authentication and digital signature system using image recognition, characterized in that the image itself is a Photoshop image produced by the user himself. The method of claim 13, wherein the database, Personal authentication and digital signature method using image recognition, characterized in that the SIFT feature points of a predetermined number of images for each scene is pre-extracted and stored. The method of claim 13, wherein the similarity between SIFT feature vectors is: Personal identification and digital signature method using image recognition, which is defined by the distance of the vector space, and the SIFT feature vectors which are very close to a certain level are determined to correspond to each other. The method of claim 13, wherein the mobile terminal, Personal authentication and digital signature method using image recognition, characterized in that to download the program necessary for the operation of the authentication image acquisition and transmission on the terminal screen of the mobile communication network connected to a specific server of the mobile communication provider. The method of claim 13 or 17, wherein the mobile terminal, Personal authentication and digital signature using video recognition, which can charge cyber money for providing information according to authentication video through mobile communication network, or perform cyber money charging after authentication process through wired network of PC Way. The method of claim 13 or 17, wherein the mobile terminal, Personal authentication and digital signature method using image recognition, characterized by storing the user's own unique image or owned in the form of a picture card and used for authentication along with a password.

Images