TW201405445A - True face recognition system and method based on dual camera - Google Patents

Abstract

The invention provided a true face recognition system and method based on dual camera, which includes two cameras, an image detection unit, a difference calculation unit, and a true human face judgment unit. These cameras are disposed and spaced apart from each other for photographing a to-be-identified object that stands in front of the cameras, so as to generate two to-be-detected images corresponding to the cameras. The image detection unit detects a facial image from these to-be-detected images, respectively. The difference calculation unit is coupled to calculate the difference of these facial images so as to output a facial difference value. The true human face judgment unit is provided to judge whether the facial images are obtained from a true human face or a photo according to the comparison result of the facial difference value and a threshold value. There is also disclosed a true face recognition method based on dual camera.

Description

Real face recognition system based on dual camera and method thereof

The present invention relates to a face recognition system and method thereof, and more particularly to a real face recognition system based on a dual camera and a method thereof.

At present, more and more TVs and various electronic display products have begun to apply face recognition technology. However, the existing face recognition technology applied to televisions and various electronic display products has the following disadvantages: first, it is easy to spoof through photos, thereby invalidating the recognition function; secondly, using infrared rays, multiple biometric identification combinations, etc. Ways to avoid similar problems with photo-spoofing must be costly, which is not conducive to the promotion of face recognition technology. Moreover, the method of avoiding similar problems with photo-spoofing due to the use of infrared rays, multiple biometric identification combinations, etc. needs to be unique. Infrared support or corresponding feature acquisition devices, which are often not highly compatible.

Therefore, in view of the above shortcomings of the existing face recognition technology, it is necessary to seek a solution.

Accordingly, it is an object of the present invention to provide a real face recognition system based on a dual camera.

Therefore, the present invention is based on a real camera recognition system of a dual camera, comprising two cameras, an image detecting unit, a difference calculating unit, and a real face determining unit.

The cameras are spaced apart from each other for shooting at the same An object to be identified in front of the camera to generate two images to be detected corresponding to the cameras, respectively.

The image detecting unit is coupled to the cameras for detecting a facial image from the waiting detection image.

The difference calculation unit is coupled to the image detection unit for calculating a difference of the facial images to output a facial difference value.

The real face determination unit is coupled to the difference calculation unit for determining that the facial images are from a real face or a photo according to the comparison result of the facial difference value and a threshold value.

Another object of the present invention is to provide a real face recognition method based on a dual camera.

Thus, the present invention is based on a real face recognition method for a dual camera, and includes the following steps.

First, an object to be recognized located in front of the cameras is photographed by two cameras arranged at a distance from each other to generate two images to be detected corresponding to the cameras, respectively.

Next, an image detecting unit detects a face image from the waiting detection image.

Then, using a difference calculation unit, the difference of the facial images is calculated to output a facial difference value.

Then, using a real face determination unit, based on the comparison result of the facial difference value and a threshold value, it is determined that the facial images are from a real face or a photo.

The above and other technical contents, features and advantages of the present invention will be apparent from the following detailed description of the preferred embodiments.

Referring to FIG. 1 and FIG. 2, the present invention is based on a preferred embodiment of a dual-camera real face recognition system, including two cameras (including a left camera 28 and a right camera 29), an image detecting unit 21, and a difference calculating unit 22. A real face determination unit 23 and a face recognition unit 24. In the preferred embodiment, the cameras 1 are spaced apart from one another and are used to capture an object 9 to be identified located in front of a range of capture angles of the cameras 28, 29 to produce two separate images. Corresponding to the images to be detected of the cameras 28, 29.

For example, as shown in FIG. 2, the cameras 28, 29, the image detecting unit 21, the difference calculating unit 22, the real face determining unit 23, and the face recognizing unit 24 of the present invention can be provided in various electronic functions having display functions. In a digital product, such as a digital television, an advertising machine, a personal computer, a tablet computer, an access control system, etc. (such as installed in the television housing 20 shown in FIG. 2), wherein the cameras 28, 29 are separated by a distance d , That is, the imaginary connecting line between the cameras 28 and 29 is a predetermined baseline of length d , and the oblique angles of the shooting directions of the cameras 28 and 29 and the preset baseline are θ ₁ and θ ₂ , respectively.

The image detecting unit 21 is coupled to the cameras 28 and 29 for detecting a facial image from the waiting detection image.

In the preferred embodiment, the facial image is generated by the image detecting unit 21 via the following pre-processing program. First, the image detecting unit 21 respectively detects a skin color region located in the image to be detected, and may Switching with the skin color application environment, for example, using a skin filter, but is not limited thereto. Next, the image detecting unit 21 performs eye feature detection from the skin color region of the waiting detection image. For example, the method of detecting the eye feature may employ Haar feature matching, but is not limited thereto. In order to extract more accurate facial images, it can also be supplemented by, for example, Principal Component Analysis, Modular Feature Extraction, Active Appearance Search Algorithm (Active Appearance) Models Search Algorithm) and so on. Since the above methods are well-known techniques, the detailed implementation thereof is well known to those skilled in the art and will not be described herein.

Therefore, the image detecting unit 21 can detect two facial images having skin color regions each having an eye feature by the above-described method.

The difference calculation unit 22 is coupled to the image detecting unit 21 for calculating a difference of the facial images to output a facial image difference value.

In the preferred embodiment, the facial image difference value is calculated by calculating an Euclidean Distance or a Mahalanobis Distance of the facial images.

The real face determination unit 23 is coupled to the difference calculation unit 22, and is configured to determine that the facial images are from real faces or photos according to the comparison result of the facial image difference value and a threshold value. .

In the preferred embodiment, the threshold is generated during the development and production phase of the dual camera based real face recognition system of the present invention. The threshold is calculated as follows.

First, in the R&D production phase, two cameras, similar to the cameras 28, 29, are used to target different real faces, so that they can be photographed within the camera-capable range of the two cameras (preferably, the different real faces are located one after another). The central fixed point position (ie, θ ₁ = θ ₂ ) is taken as shown in FIG. ₂ ; then, for the two images corresponding to each real face, an image detecting unit similar to the image detecting unit 21 is used. Perform image preprocessing on the two images separately (ie, detecting the skin color region first, and then performing eye feature detection from the skin color region) to obtain the processed face image; and then using the difference calculation unit 22 The difference calculation unit calculates the difference value between the two images of each group (such as Euclidean distance or horse distance). Accordingly, this can develop the production stage, using a threshold value operation unit (not shown), by statistical method, the minimum difference value is obtained from the difference value X _min of the face image in the array.

Then, in the R&D production stage, the two left and right cameras are used to shoot different face photos, so that they can be photographed within the range of the two cameras (preferably, the different face photos are located as shown in FIG. 2). The central fixed point position (ie, θ ₁ = θ ₂ ) is taken); then, for the two images corresponding to each face photo, the image detecting unit is used to perform image preprocessing on the two images respectively. The processed image is obtained; then the difference calculation unit is used to calculate the difference value between the two images of each group. Therefore, in this R&D production stage, the threshold value calculation unit can be used to obtain the maximum difference value Y _max from the difference value of the array photo image by statistical means.

Then, this development can production stage, using the threshold value calculating means calculates the average of the threshold value is equal to the minimum difference value X _min and Y _max of the maximum difference value.

Therefore, in the user operation phase, if the facial image difference value is greater than the When the threshold value is reached, the real face determination unit 23 determines that the image to be recognized is from a real face, and if the facial image difference value is smaller than the threshold value, the real face determination unit 23 determines that the to-be-determined The recognition image is from a photo.

The face recognition unit 24 is coupled to the real face determination unit 23, and the face recognition unit 24 is configured to use one of the facial images when the facial images are from a real face. The identity of the image to be recognized is identified.

The face recognition unit 24 adopts one selected from the group consisting of Gray Balance, Light Compensation, and Thresholding in performing normalization on the face image. .

The face recognition unit 24 performs a feature extraction process for the face image, which is selected from Principal Component Analysis, Modular Feature Extraction, and active appearance model search calculus. One of the Active Appearance Models Search Algorithms.

The face recognition unit 24 adopts a feature selected from the Euclidean Distance, the K-Nearest Neighbor, and the Support Vector Machine in performing the feature classification process for the face image. And one of the Neural Networks.

DETAILED DESCRIPTION OF THE INVENTION In the patent specification of the present invention, a preferred embodiment of the real face recognition method based on the dual camera of the present invention will be described in detail below with reference to Figs.

As shown in step 31, an image of the object to be identified located in front of the cameras 28, 29 is captured by the two cameras 28, 29 spaced apart from each other to produce two frames corresponding to the cameras 28, 29, respectively. Image to be detected.

As shown in step 32, the image detecting unit 21 detects a face image from the waiting detection image.

The image detecting unit 21 first detects a skin color region from the waiting detection image, and then performs eye feature detection from the skin color regions to detect the face image from the waiting detection image. image.

As shown in step 33, the difference calculation unit 22 calculates the difference of the facial images to output a facial image difference value.

The difference calculation unit 22 obtains the face image difference value by calculating the Euclidean distance or the horse-like distance of the facial images.

As shown in step 34, the real face determination unit 23 determines, based on the comparison result of the facial image difference value and the threshold value, that the facial images are from a real face or a photo.

If the facial image difference value is greater than the threshold value, it indicates that the image to be recognized is from a real human face, and then, as shown in step 35, the facial recognition unit 24 is from one of the facial images. , identifying the identity of the image to be identified.

Conversely, if the facial image difference value is less than the threshold value, it indicates that the image to be recognized is from a photo, and then, as shown in step 36, the preferred embodiment of the present invention is an electronic device having a display function, such as a television set. A warning message is displayed on the user interface of the digital product, such as "Invalid recognition" Wait.

In summary, the present invention can determine the difference between the facial image differences between the left and right facial images, and then determine the comparison based on the comparison between the facial image difference value and the threshold value. The face image is from a real face or a photo, so that the problem that the face recognition technology used in televisions and various electronic display products is easily spoofed through photos can be effectively avoided; moreover, since the present invention utilizes each other The two cameras are arranged at intervals to obtain the image to be detected, so the cost is low, and it is easy to apply and popularize. Moreover, since the dual camera is widely used in the three-dimensional modeling technology, the market is large and highly compatible. Sexuality, it is indeed possible to achieve the object of the present invention.

The above is only the preferred embodiment of the present invention, and the scope of the invention is not limited thereto, that is, the simple equivalent changes and modifications made by the scope of the invention and the description of the invention are All remain within the scope of the invention patent.

20‧‧‧TV housing

21‧‧‧Image detection unit

22‧‧‧Differential calculation unit

23‧‧‧Real face determination unit

24‧‧‧Face recognition unit

28‧‧‧ camera

29‧‧‧ camera

31~36‧‧‧Steps

1 is a schematic view showing a schematic diagram of a two-camera based on a dual-camera real face recognition system in which a plurality of cameras capture an object to be recognized in front of the cameras; FIG. 2 is a functional block diagram. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention is based on a preferred embodiment of a dual face camera real face recognition system; and FIG. 3 is a flow chart illustrating a preferred embodiment of the present invention based on a dual camera real face recognition method.

21‧‧‧Image detection unit

22‧‧‧Differential calculation unit

23‧‧‧Real face determination unit

24‧‧‧Face recognition unit

28‧‧‧ camera

29‧‧‧ camera

Claims (10)

A real face recognition system based on a dual camera, comprising: two cameras arranged at intervals to capture an object to be identified located in front of the cameras to generate two images respectively corresponding to the object to be identified An image detecting unit coupled to the cameras for detecting a facial image from the waiting detection image; a difference calculating unit coupled to the image detecting unit Calculating the difference of the facial images to output a facial image difference value; and a real human face determining unit coupled to the difference calculating unit for using the facial image difference value and the As a result of the comparison of the threshold values, it is determined that the facial images are from real faces or photos. The real face recognition system for a dual camera according to claim 1, wherein the image detecting unit first detects a skin color region from the waiting detection image, and then separately performs an eye from the skin color regions. Part feature detection to detect the face image from the waiting detection image. The real face recognition system of the dual camera according to claim 1, wherein the difference calculation unit calculates the face distance or the horse distance of the face images to obtain the difference of the face image. value. According to the real face recognition system of the dual camera of claim 1, wherein if the facial image difference value is greater than the threshold value, the real face determination unit determines that the to-be-identified object system is from the real people a face, and if the facial image difference value is less than the threshold value, the real face determination unit determines that the to-be-identified object system is from a photo. The real face recognition system of the dual camera according to claim 1, further comprising a face recognition unit coupled to the real face determination unit, wherein the facial images are from a real face The face recognition unit is configured to identify the identity of the object to be identified from one of the facial images. A real face recognition method based on a dual camera, comprising the steps of: (A) capturing two objects to be recognized in front of the cameras by using two cameras arranged at intervals with each other to generate two frames respectively corresponding to the cameras (B) using an image detecting unit to detect a face image from the waiting detection image; (C) using a difference calculation unit to calculate the face images a difference between the images to output a facial image difference value; and (D) using a real human face determining unit to determine the facial image based on the comparison between the facial image difference value and a threshold value It comes from real faces or photos. The dual-camera-based real face recognition method according to claim 6, wherein in step (B), the image detecting unit first detects a skin color region from the waiting detection image, and then Eye feature detection is performed in each of the skin color regions to detect the face image from the waiting detection image. The dual-camera-based real face recognition method according to claim 6, wherein in the step (C), the real face recognition module calculates the Euclidean distance of the facial images or Horse distance to get the facial image difference value. The dual-camera-based real face recognition method according to claim 6, wherein in the step (D), if the facial image difference value is greater than the threshold value, the real face determination unit determines The system to be identified is derived from a real face, and if the facial image difference value is less than the threshold value, the real face determination unit determines that the to-be-identified object system is from a photo. The dual-camera-based real face recognition method according to claim 6 of the patent application further includes a step (E) after the step (D), when the facial images are from a real face And identifying, by using a face recognition unit, the identity of the object to be identified from one of the facial images.