Dual-camera face recognition device and method

FIELD OF THE INVENTION

p-0002The present invention relates to a biometric recognition field, and especially, to a face recognition device and method, particularly relates to a dual cameral face recognition device and method.

BACKGROUND OF THE INVENTION

p-0003Biometric recognition technology is known as one of the top ten technologies that bring revolutionary changes in human society. As the most convenient and secure identification recognition technology, the biometric recognition technology recognizes the human beings per se without any mark outside of the human beings. The biometric recognition technology performs identification by using physical characteristics and behavior of human, mainly including fingerprint recognition, face recognition, iris recognition, gait recognition and the like. Among other things, the face recognition is currently popular in the biometric recognition field. As compared with fingerprint recognition technology applied widely, it has many significant advantages, for example, it is intuitive, convenient, non-contact, friendliness and of high user acceptance and so on.

p-0004The existing face recognition technology can be divided into two-dimensional face recognition and three-dimensional face recognition. Three-dimensional face recognition is based on three-dimensional face images, however, there are several disadvantages such as the complexity of a capturing system and face reconstruction algorithms. For example, as seen mentioned in Application No. “CN 1595280A” and titled as “THREE-DIMENSIONAL HUMAN FACE REBUILDING SYSTEM ON ROTATING PLATFORM”, No. “CN 1315092C” and titled as “THREE DIMENSION FACE RECOGNITION METHOD BASED ON POLAR SPECTRUM IMAGE”, the existing three-dimensional face recognition is expensive and restrictive.

p-0005Two-dimensional face recognition is based on a single planar face image. In general, a two-dimensional face recognition method includes steps of capturing a planar face image by means of one camera, performing face detection, eye location and feature extraction, then comparing with a template library and making recognition decision. However, the recognition performance is poor since a recognition rate for a single face plane image captured by one camera is actually affected by ambient lights, capturing angels, postures, face expressions and the other factors.

SUMMARY OF THE INVENTION

p-0006In order to solve the above problem, a dual-camera face recognition device and method is provided in one or more embodiments of the present invention, which may significantly improve the recognition performance for the existing two-dimensional face recognition.

p-0007According to an aspect of the present invention, a dual-camera face recognition device, comprising: a first camera, for capturing a first face image; a second camera, arranged at a position different from the first camera, for capturing a second face image; a face recognition processing unit, for receiving and recognizing the first face image captured by the first camera and obtaining a first candidate set, and for receiving and recognizing the second face image captured by the second camera and obtaining a second candidate set, and for selecting from the first candidate set and the second candidate set a candidate object, whose similarity complies with a predetermined rule, as a recognition result.

p-0008Additionally, the dual-camera face recognition device further comprises an infrared fixed active light source, arranged behind the first camera and the second camera, for serving as a light source for capturing face images. As an example of the present invention, the dual-camera face recognition device further comprises an infrared transmission filter, arranged in front of the first camera, for filtering out visible lights such that a black and white face image is captured by the first camera; and an infrared cut filter, arranged in front of the second camera, for filtering out infrared lights such that a color face image is captured by the second camera. Preferably, the dual-camera face recognition device further comprises a display unit, for receiving and displaying the color face image captured by the second camera. As an example of the present invention, the dual-camera face recognition device further comprises infrared transmission filters, arranged in front of the first camera and the second camera respectively, for filtering out visible lights such that the first camera and the second camera respectively captures a black and white face image.

p-0009According to a first embodiment of the present invention, the face recognition processing unit is configured to compute a sum of a similarity of a candidate object in the first candidate set with respect to the first face image captured by the first camera and a similarity of the same candidate object in the second candidate with respect to the second face image capture by the second camera if any, for each of the candidate objects in the first candidate set, so as to obtain sums of similarities, determine whether a maximum value of the sums of similarities is greater than a predetermined threshold, and select a candidate object corresponding to the maximum value as a recognition result when the maximum value is greater than the predetermined threshold.

p-0010According to a second embodiment of the present invention, the face recognition processing unit is configured to determine whether a maximum value among similarities of all candidate objects in the first candidate set and the second candidate set is greater than a predetermined threshold, and selects a candidate object corresponding to the maximum value as a recognition result when the maximum value is greater than the predetermined threshold.

p-0011According to a third embodiment of the present invention, the face recognition processing unit is configured to select from the first candidate set a candidate object whose similarity is greater than a first predetermined threshold, and select from the second candidate set a candidate object whose similarity is greater than a second predetermined threshold, so as to form a third candidate set, compute a sum of a similarity of a candidate object in the third candidate set with respect to the face image captured by the first camera and a similarity of the same candidate object in the third candidate set with respect to the face image captured by the second camera if any, for all of the candidate objects in the third candidate set, so as to obtain sums of similarities, determine whether a maximum value of the sums of similarities is greater than a third predetermined threshold, and select a candidate object corresponding to the maximum value as a recognition result when the maximum value is greater than the third predetermined threshold.

p-0012According to another aspect of the present invention, a dual-camera face recognition method is provided, which comprises steps of capturing a first face image by a first camera; capturing a second face image by a second camera arranged at a position different from the first camera; recognizing the first face image captured by the first camera and obtaining a first candidate set, and recognizing the second face image captured by the second camera and obtaining a second candidate set; and selecting from the first candidate set and the second candidate set a candidate object, whose similarity complies with a predetermined rule, as a recognition result.

p-0013In addition, the dual-camera face recognition method further comprises step of providing each of the first camera and the second camera with an infrared fixed active light source as a light source for capturing face images. As an example of the present invention, the step of capturing a first face image by the first camera further comprises steps of: filtering out visible lights by an infrared transmission filter arranged in front of the first camera and capturing a black and white image by the first camera; and the step of capturing a second face image by the second camera arranged at a position different from the first camera further comprises steps of: filtering out infrared lights by an infrared cut filter arranged in front of the second camera and capturing a color face image by the second camera. Preferably, the dual-camera face recognition method further comprises a step of displaying on a display unit the color face image captured by the second camera. As another example of the present invention, the steps of capturing the first face image by the first camera and capturing the second face image by the second camera arranged at a position different from the first camera further comprise steps of: filtering out visible lights by infrared transmission filters, which are arranged in front of the first camera and the second camera, respectively, and capturing black and white face images by the first camera and the second camera.

p-0014According to a first embodiment of the present invention, the step of selecting from the first candidate set and the second candidate set a candidate object, whose similarity complies with a predetermined rule, as a recognition result further comprises steps of: computing a sum of a similarity of candidate object in the first candidate set with respect to the first face image captured by the first camera and a similarity of the same candidate object in the second candidate with respect to the second face image capture by the second camera if any, for each of the candidate objects in the first candidate set, so as to obtain sums of similarities; determining whether a maximum value among the sums of similarities is greater than a predetermined threshold; and selecting a candidate object corresponding to the maximum value as a recognition result when the maximum value is greater than the predetermined threshold.

p-0015According to a second embodiment of the present invention, the step of selecting from the first candidate set and the second candidate set a candidate object, whose similarity complies with a predetermined rule, as a recognition result further comprises steps of: determining whether a maximum value among similarities of all candidate objects in the first candidate set and the second candidate set is greater than a predetermined threshold; and selecting a candidate object corresponding to the maximum value as a recognition result when the maximum value is greater than the predetermined threshold.

p-0016According to a third embodiment of the present invention, the step of selecting from the first candidate set and the second candidate set a candidate object, whose similarity complies with a predetermined rule, as a recognition result further comprises steps of: selecting from the first candidate set a first candidate object whose similarity is greater than a first predetermined threshold and selecting from the second candidate set a second candidate object whose similarity is greater than a second predetermined threshold, so as to form a third candidate set; computing a sum of a similarity of a candidate object in the third candidate set with the first face image captured by the first camera and a similarity of the same candidate object in the third candidate set with respect to the second face image capture by the second camera if any, for each of the candidate objects in the third candidate set so as to obtain sums of similarities; determining whether a maximum value of the sums of similarities is greater than a third predetermined threshold; and selecting a candidate object corresponding to the maximum value as a recognition result when the maximum value is greater than the third predetermined threshold.

p-0017On the basis of the existing face capturing usually employing a single camera, the face recognition device and method according to the present invention employs dual cameras to perform face capturing and face recognition and then performs a selection on two candidate sets based on a predetermined rule. The present invention is easily implemented and may significantly improve recognition performance of the existing two-dimensional face recognition. In addition, the face recognition device according to the present invention may display a color image on a display unit, which may bring good user experience.

BRIEF DESCRIPTION OF THE DRAWINGS

p-0018For a more complete understanding of the present invention and its advantages, detailed embodiments of the present invention will be described with reference to the drawings. Please be noted that similar reference signs represent similar components throughout the drawings, in which:

p-0019<figref idrefs="DRAWINGS">FIG. 1</figref> is a schematic diagram of a dual-camera face recognition device according to an embodiment of the present invention;

p-0020<figref idrefs="DRAWINGS">FIG. 2</figref> is a schematic diagram of a dual-camera face recognition device, which employs an infrared fixed active light source, according to an embodiment of the present invention;

p-0021<figref idrefs="DRAWINGS">FIG. 3</figref> is a flowchart of a dual-camera face recognition method according to a first embodiment of the present invention;

p-0022<figref idrefs="DRAWINGS">FIG. 4</figref> is a flowchart of a dual-camera face recognition method according to a second embodiment of the present invention; and

p-0023<figref idrefs="DRAWINGS">FIG. 5</figref> is a flowchart of a dual-camera face recognition method according to a third embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

p-0024<figref idrefs="DRAWINGS">FIG. 1</figref> is a schematic diagram of a dual-camera face recognition device according to an embodiment of the present invention. A first camera <b>1</b> and a second camera <b>2</b> are respectively placed at different positions, which are suitable for capturing face images, so as to capture face images. For example, the first camera <b>1</b> and the second camera <b>2</b> may be placed at a certain distance in the horizontal direction or in the vertical direction. Alternatively, the cameras may be placed at other positions, which may facilitate capturing face images, depending on placement conditions, appearances, invisibility and the other factors. Preferably, they may be placed in the same height with a certain interval, such as 10 cm. The first camera <b>1</b> and the second camera <b>2</b> may be electrically connected to a face recognition processing unit <b>101</b> respectively, for signal transmission to the face recognition processing unit <b>101</b>. The face recognition processing unit <b>101</b> may employ a PC platform or an embedded platform, such as DSP, ARM and the like. When a person stands within a certain range in front of the first camera <b>1</b> and the second camera <b>2</b>, the first camera <b>1</b> and the second camera <b>2</b> will respectively capture face images and transmit them to the face recognition processing unit <b>101</b>. The face recognition processing unit <b>101</b> then performs face recognition processing using image information captured by the two cameras, so as to enhance light, gesture and face expression adaptability of the common two-dimensional face recognition system.

p-0025In order to enhance light adaptability of the face recognition device according to the present invention, the face recognition device according to the present invention may be provided with an active light source. <figref idrefs="DRAWINGS">FIG. 2</figref> is a schematic diagram of a dual-camera face recognition device according to an embodiment of the present invention. In this embodiment, an infrared fixed active light source serves as a light source for capturing a face image, so as to enhance light adaptability of the common two-dimensional face recognition device. In the dual-camera face recognition device according to the embodiment, the infrared fixed active light source is composed of a plurality of infrared light-emitting diodes that are uniformly distributed, and is of a central wavelength of 700-2000 nm, preferably as 850 nm. This light source may be arranged behind the first camera <b>1</b> and the second camera <b>2</b>. It should be appreciated that light sources of the same functions may be employed, other than that illustrated in this embodiment.

p-0026It should be appreciated that the two cameras may capture the same type of face images. For example, infrared transmission filters may be arranged in front of each camera respectively for filtering out visible lights so as to capture a black and white face image, or infrared cut filters may be arranged in front of each camera for filtering out infrared lights so as to capture a color face image.

p-0027Furthermore, in order to improve the recognition efficiency, the two cameras may be allowed to capture different type of face images. For example, one camera captures a black and white image, and the other captures a color image. To prevent lens of a camera from being affected by the ambient light, preferably, an infrared transmission filter may be installed in front of lens of one camera, which is capable of filtering out visible lights while with infrared lights penetrating, or a light filter may be directly coated onto the lens, so as to directly capture a black and white image. Moreover, in order to prevent lens of a camera from being affected by the infrared lights, preferably, an infrared cut filter may be installed in front of lens of the other camera, which is capable of filtering out infrared lights while with visible lights penetrating. Or a light filter may be directly coated onto the lens, so as to capture a color image. CMOS or CCD may be used as an image sensor of the cameras. Of course, the present invention is not limited to the above embodiment, and those skilled in the art may perform various settings and modifications on natures of images captured by the cameras depending on various applications, alternatively, additional devices may be added for capturing images of specific natures, which all fall into the scope of the present invention.

p-0028In addition, in order to bring good user experience while capturing a black and white image, the face recognition device according to the present invention may further comprise a display unit (not shown). When either of the first camera and the second camera captures a color face image, the display unit may receive and display the color face image captured by the camera.

p-0029<figref idrefs="DRAWINGS">FIG. 3</figref> is a flowchart of a dual-camera face recognition method according to a first embodiment of the present invention. At steps <b>301</b> and <b>302</b>, a first camera and a second camera respectively start to perform face recognition. At steps <b>303</b> and <b>304</b>, the first camera and the second camera respectively capture a face image. A face recognition processing unit recognizes the face image captured by the first camera (step <b>305</b>) and the face image captured by the second camera (step <b>306</b>) so as to obtain two face recognition candidate sets A and B respectively and respective similarity corresponding to each candidate object in the candidate sets A and B. Suppose numbers of candidate objects in the sets A and B are m and n, respectively. At step <b>307</b>, when there are the same candidate object in the candidate set B, for each of the candidate objects in the candidate set A, a sum of a similarity of candidate object in the candidate set A with respect to the face image captured by the first camera and a similarity of the same candidate object in the candidate set B with respect to the face image capture by the second camera are computed, so as to obtain sums of similarities, and then the sums of similarities are sorted in a descending order so as to obtain a combined similarity ranking for the candidate objects. At step <b>308</b>, whether the maximum value of the sums of the similarities (i.e., a sum of similarity with highest level in the similarity combined ranking) is greater than a predetermined threshold is determined. When the maximum value is greater than the predetermined threshold, a candidate object corresponding to the maximum value is selected as a recognition result, as shown at step <b>309</b>.

p-0030At steps <b>305</b> and <b>306</b>, the face recognition processing unit may perform face recognition based on various recognition algorithms. In addition, the predetermined threshold is a value obtained by repeated tests based on values of similarity, recognition rates and the other factors.

p-0031<figref idrefs="DRAWINGS">FIG. 4</figref> is a flowchart of a dual-camera face recognition method according to a second embodiment of the present invention. At steps <b>401</b> and <b>402</b>, a first camera and a second camera respectively start to perform face recognition. At steps <b>403</b> and <b>404</b>, the first camera and the second camera respectively capture a face image. A face recognition processing unit recognizes the face image captured by the first camera (step <b>405</b>) and the face image captured by the second camera (step <b>406</b>) so as to obtain two face recognition candidate sets A and B and similarities corresponding to each of candidate objects in the candidate sets A and B. Suppose numbers of candidate objects in the sets A and B are m and n, respectively. At step <b>407</b>, it is determined whether a maximum value of the similarities for all the m+n candidate objects in the candidate sets A and B is greater than a predetermined threshold. When the maximum value is greater than the predetermined threshold, a candidate object corresponding to the maximum value is selected as a recognition result, as shown at step <b>408</b>.

p-0032<figref idrefs="DRAWINGS">FIG. 5</figref> is a flowchart of a dual-camera face recognition method according to a third embodiment of the present invention. According to the dual-camera face recognition method according to this embodiment, a face recognition processing unit <b>101</b> firstly selects candidate objects of greater similarities with respect to a face image captured by a first camera in the candidate set A and candidate objects of greater similarities with respect to a face image captured by a second camera in the candidate set B so as to remain these candidate objects in the sets, and then selects a candidate object as a recognition result from the remained candidate objects based on a predetermined rule. At steps <b>501</b> and <b>502</b>, a first camera and a second camera respectively start to perform face recognition. At steps <b>503</b> and <b>504</b>, the first camera and the second camera respectively capture a face image. A face recognition processing unit recognizes the face image captured by the first camera (step <b>505</b>) and the face image captured by the second camera (step <b>506</b>) so as to obtain two face recognition candidate sets A and B and similarities corresponding to each of the candidate objects in the candidate sets A and B.

p-0033Suppose numbers of candidate objects in the sets A and B are m and n, respectively. At step <b>507</b>, a candidate object whose similarity is greater than a first predetermined threshold is selected from the candidate set A and a candidate object whose similarity is greater than a second predetermined threshold is selected from the candidate set B, so as to form a new candidate set C. Suppose the number of the candidate objects selected from the set A is m<b>1</b> and m<b>1</b>≦m, and the number of the candidate objects selected from the set B is n<b>1</b> and n<b>1</b>≦n. At step <b>508</b>, for each of the candidate objects in set C, a sum of a similarity of the candidate object of the m<b>1</b> candidate objects with respect to the face image captured by the first camera and a similarity of the same candidate object of the n<b>1</b> candidate objects with respect to the face image capture by the second camera is computed, so as to obtain sums of similarities, and then the sums of similarities are sorted in a descending order so as to obtain similarity combined ranking for the candidate objects. At step <b>509</b>, a maximum value in the similarity combined ranking, i.e. whether the maximum value of the sums of the similarities is greater than a third predetermined threshold is determined. When the maximum value is greater than the third predetermined threshold, a candidate object corresponding to the maximum value is selected as a recognition result. At steps <b>507</b> and <b>509</b>, the first, second and third predetermined thresholds are values obtained by repeated tests based on values of similarity, recognition rates and the other factors. The first predetermined threshold may be equal to the second predetermined threshold, which depends on different properties of images captured by the cameras. Of course, they may be different values.

p-0034On the basis of the existing face capturing usually employing a single camera, the present invention proposes a dual-camera face recognition device, which can significantly improve recognition performance of the existing two-dimensional face recognition. Additionally, in order to significantly prevent the recognition operation from ambient lights, the existing face recognition system generally captures black and white face images for display and recognition, which however brings poor user experience. Therefore, the present invention is also benefit in bringing good user experience while capturing black and white face images.

p-0035The foregoing description gives only the preferred embodiments of the present invention. It is obvious that any modification, substitution, improvement or like falls into the spirit and principle of the present invention defined by Claims attached.