Summary of the invention
In view of this, the object of this invention is to provide a kind of remote multi-modal biological characteristic recognition methods and system thereof, its multi-modal biological characteristic recognition technology utilizing iris mode and face mode to combine, build high-precision optical system platform, adopt intelligent human-machine interaction mode, more traditional single living things feature recognition, not only can realize maximizing favourable factors and minimizing unfavourable ones between each feature, have complementary advantages, and can while remote identification, the further accuracy improving identification, realize remote, high-resolution, multi-modal biological characteristic identification at a high speed, the product use sense being conducive to improving user is subject to, be of great practical significance.
For this reason, the invention provides a kind of remote multi-modal biological characteristic recognition methods, comprise step:
The first step: the depth image of shooting, collecting user, obtains the depth information of user;
Second step: according to the depth information of obtained user, obtains the depth distance of user position;
3rd step: judge whether the depth distance of user position is positioned within the distance range of user preset, if, then send default facial image shooting to user and start information, and continue to perform step the four step, if not, when described depth distance is less than the minimum threshold values of user preset distance range or when being greater than maximum threshold values, correspondence sends default facial image shooting and prepares information respectively;
4th step: the direct picture of shooting, collecting user;
5th step: Face datection is carried out to the direct picture of the user of captured collection, then locating segmentation obtains the facial image in user's direct picture;
6th step: whether the image quality parameter that the facial image that judgement obtains has is positioned at the number range of default quality of human face image parameter, if so, is then stored in real time by described facial image, if not, returns and performs step the four step;
7th step: send the near infrared light needed for iris imaging to user, and the face subregion image of shooting, collecting user, and segmentation obtains human eye area image from the image of described face subregion; 8th step: human eye detection is carried out to the human eye area image of obtained user, then locating segmentation obtains the human eye topography in user's human eye area image;
9th step: judge whether the image quality parameter that the human eye topography obtained has is positioned at default eye image mass parameter number range, if so, then described human eye topography is stored in real time, if not, returns and performs step the seven step;
Tenth step: the iris region in human eye topography is positioned, then extract face characteristic that the facial image of active user that stores has and extract the iris feature that in human eye topography, iris region has, and the face characteristic that has respectively of the face characteristic this user had and iris feature and the multiple validated users prestored and iris feature information compare and mates, complete identifying.
Wherein, in the described first step and second step, gathered the depth image of shooting, collecting user by depth camera, obtain the depth distance of user position.
Wherein, in described 4th step, the direct picture of user is gathered by face video camera;
Described face video camera is the lightfield camera that autozoom video camera or numeral are heavily focused.
Wherein, in described 7th step, send near infrared light needed for iris imaging by the light source that is made up of LED array to user, and by the face subregion image of iris video camera shooting, collecting user;
Described iris video camera is the lightfield camera that autozoom video camera or numeral are heavily focused.
In addition, present invention also offers a kind of remote multi-modal biological characteristic recognition system, comprising:
Depth camera, for the depth image of shooting, collecting user, obtains the depth information of user;
Central processing control module, be connected with depth camera, controlling depth video camera startup optimization, according to the depth information of the user that depth camera obtains, obtain the depth distance of user position, and judge whether the depth distance of user position is positioned within the distance range of user preset, if, then send default facial image shooting to user and start information, and trigger operation face video camera, if not, when described depth distance is less than the minimum threshold values of user preset distance range or when being greater than maximum threshold values, correspondence sends default facial image shooting and prepares information respectively,
Face video camera, for the direct picture of shooting, collecting user;
Described central processing control module is also connected with face video camera, direct picture for the user to the collection of face shot by camera carries out Face datection, then locating segmentation obtains the facial image in user's direct picture, and judge whether image quality parameter that the facial image that obtains has is positioned at the number range of default quality of human face image parameter, if, then described facial image real-time storage is stored recognition unit to preset data, and trigger operation light source module and iris video camera, if not, continue the direct picture controlling face video camera shooting, collecting user,
Light source module, for sending the near infrared light needed for iris imaging to user;
Iris video camera, for the face subregion image of shooting, collecting user, and segmentation obtains human eye area image from the image of described face subregion;
Described central processing control module is also connected with iris video camera, for controlling iris video camera startup optimization, and to iris video camera gather the user of acquisition human eye area image carry out human eye detection, then locating segmentation obtains the human eye topography in user's human eye area image, and judge whether the image quality parameter that the human eye topography obtained has is positioned at default eye image mass parameter number range, if, then described human eye topography real-time storage is stored recognition unit to preset data, if not, continue controlling run light source module and iris video camera,
Preset data stores recognition unit, be connected with described central processing control module, the user's facial image sent for central processing control module described in real-time storage and human eye topography, and extract face characteristic that the facial image of active user that stores has and the iris feature that human eye topography iris region has, and the face characteristic that has respectively of the face characteristic this user to be had and iris feature and the multiple validated users prestored and iris feature information compare and mates, complete identifying.
Wherein, comprise system shell, the lower end, front of described system shell has described central processing control module, described light source module is provided with directly over described central processing control module, be provided with display screen directly over described light source module, the both sides up and down of described display screen are respectively arranged with a described iris video camera;
Described two iris video cameras in distributing up and down, are respectively used to the face subregion image of the user gathering different height in the vertical direction.
The left side of described display screen is provided with described face video camera, and the right side of described display screen is provided with described depth camera, and the top of described depth camera has loud speaker.
Wherein, described depth camera comprises any one in time-of-flight method TOF camera, structured light depth camera and laser scanning depth camera; Described face video camera is that autozoom video camera or numeral are heavily focused lightfield camera; Described iris video camera is that autozoom video camera or numeral are heavily focused lightfield camera.
Wherein, when described face video camera and iris video camera are autozoom video camera, the autozoom mode that described face video camera and iris video camera are taked is specially: realize zoom by the camera lens photocentre of mobile autozoom video camera, or is moved by the transducer controlling autozoom video camera and realize zoom.
Wherein, described preset data stores identification form is cloud platform or the remote server with data storage function.
Wherein, described central processing control module comprises autozoom and controls submodule and numeral and heavily to focus control submodule, wherein:
Autozoom controls submodule, be connected with iris video camera with face video camera respectively, for when described face video camera and iris video camera adopt autozoom video camera, according to the depth distance value of the user present position that depth camera obtains, determine the focusing position of described face video camera and iris video camera, control described face video camera and iris camera driver and change the camera lens of video camera and the relative position of transducer, realize auto-focusing, obtain the shape library image of face or iris.
Numeral is heavily focused control submodule, be connected with iris video camera with face video camera respectively, for when described face video camera and iris video camera adopt lightfield camera, control described face video camera and the original face of iris camera acquisition and eye image and transfer to central processing control module, then, according to the depth distance of the user present position that depth camera obtains, determine the focusing position of described face video camera and iris video camera, then parameter of heavily focusing is determined, the heavy focus operation of numeral is carried out to original face and face subregion image, obtain the shape library image of face or iris.
From above technical scheme provided by the invention, compared with prior art, the invention provides a kind of remote multi-modal biological characteristic recognition methods and system thereof, its multi-modal biological characteristic recognition technology utilizing iris mode and face mode to combine, break through single mode feature in the scope of application, the limitation of the aspect such as accuracy of identification and safety anti-fake, break through existing imaging technique and obtain the technical bottleneck in remote biometric image, build high-precision optical system platform, adopt intelligent human-machine interaction mode, more traditional single living things feature recognition, not only can realize maximizing favourable factors and minimizing unfavourable ones between each feature, have complementary advantages, and can while remote identification, the further accuracy improving identification, realize remote, high-resolution, multi-modal biological characteristic identification at a high speed, the product use sense being conducive to improving user is subject to, be of great practical significance.
Embodiment
In order to make those skilled in the art person understand the present invention program better, below in conjunction with drawings and embodiments, the present invention is described in further detail.
Fig. 1 is the flow chart of a kind of remote multi-modal biological characteristic recognition methods provided by the invention;
See Fig. 1, the remote multi-modal biological characteristic recognition methods of one provided by the invention, comprises the following steps:
Step S101: the depth image of shooting, collecting user, obtains the depth information (three-dimensional feature information of user) of user;
Step S102: according to the depth information of obtained user, obtains the depth distance (distance namely between user position and spot for photography) of user position;
For the present invention, for step S101 and S102, in specific implementation, can by the depth image of depth camera shooting, collecting user, and export the depth distance of user position, particular by multi views anaglyph, photometric stereo, colourity forming process or defocus the existing depth computing methods based on image such as deduction method and obtain depth distance between user position and spot for photography (as depth camera infield).
Step S103: judge whether the depth distance of user position is positioned within the distance range of user preset, if, then send default facial image shooting to user and start information, and continue to perform step S104, if not, when described depth distance is less than the minimum threshold values of user preset distance range or when being greater than maximum threshold values, correspondence sends default facial image shooting and prepares information respectively;
For the present invention, in step s 103, in specific implementation, the shooting of described facial image start information can according to user need arrange in advance, can be such as one section of sound of " movement please be stop, standing still and look squarely front " for the content of played pre-recorded.
For the present invention, in step s 103, in specific implementation, when described depth distance is less than the minimum threshold values of user preset distance range, the default described facial image shooting that correspondence sends prepares one section of sound that information can be " please moving backward " for the content prerecorded; When described depth distance is greater than the maximum threshold values of user preset distance range, the facial image the preset shooting that correspondence sends prepares one section of sound that information can be " please move forward " for the content prerecorded.
Step S104: the direct picture of shooting, collecting user;
For the present invention, in step S104, can be carried out the direct picture of shooting, collecting user by face video camera, described face video camera can adopt the video camera of COMS or ccd imaging sensor.Described face video camera is preferably and adopts autozoom video camera or can realize digital lightfield camera of heavily focusing.
Step S105: Face datection is carried out to the direct picture of the user of captured collection, then locating segmentation obtains the facial image in user's direct picture;
For the present invention, it should be noted that, Face datection and location belong to specific algorithm part, and in specific implementation, the present invention can utilize the result of the Face datection of existing maturation and location scheduling algorithm to be partitioned into facial image.
In specific implementation, such as can adopt and carry out Face datection based on haar (Ha Er) classifier algorithm, (SupervisedDescentMethod (having the gradient descent method of supervision) algorithm carries out face key point location to utilize SDM.
Step S106: whether the image quality parameter that the facial image that judgement obtains has is positioned at the number range (namely carrying out quality of human face image evaluation) of default quality of human face image parameter, if, then described facial image is stored in real time, if not, execution step S104 is returned;
For the present invention, in step s 106, described image quality parameter can comprise one or more in resolution, contrast and colour temperature.The number range of described default quality of human face image parameter, can according to user need arrange in advance.
For the present invention, in step s 106, described facial image can be stored into cloud platform, remote server or other devices with data storage function.
Step S107: send the near infrared light needed for iris imaging to user, and the face subregion image of shooting, collecting user is (compared with facial image, the image gathered around two of user when this face subregion image is iris imaging, just account for a part for facial image, the ratio specifically accounted for is relevant to the distance between iris imaging equipment and user such as iris video camera, distance is nearer, the ratio then accounted in whole facial image is less, otherwise it is larger), and Iamge Segmentation obtains human eye area image (i.e. horizontal covering two and eyebrow in face from described face subregion, the bridge of the nose is at interior one section of image),
It should be noted that, for the present invention, in step s 107, according to the nominal data between the facial image positioning result of step S105 and face video camera, iris video camera, can take to dig window shape formula from the image of face subregion, split acquisition human eye area image, digging window is exactly only interested region (referring to human eye area in this step) in the area image of protoplast face part is split.
Described nominal data refers to the pixel corresponding relation that in different cameras image, overlapped fov is interregional, step S105 can obtain the human eye key point position in facial image, according to the demarcation relation of face video camera and iris video camera, and then the relevant position of the corresponding human eye key point of face subregion image can be obtained, can by human eye area Iamge Segmentation out according to this position.
For the present invention, in step s 107, near infrared light needed for iris imaging can be sent by the light source that is made up of LED array to user.
For the present invention, in step s 107, can by the face subregion image of existing iris video camera shooting, collecting user.Described iris video camera is preferably and adopts autozoom video camera or can realize digital lightfield camera of heavily focusing.
Step S108: human eye detection is carried out to the human eye area image of obtained user, then the human eye topography in locating segmentation acquisition user human eye area image is (compared with human eye area image, this human eye topography only has right and left eyes near the eyes, does not have eyebrow, the bridge of the nose);
For the present invention, it should be noted that, human eye detection and location belong to specific algorithm part, and in specific implementation, the present invention can utilize the result of this kind of algorithms such as existing human eye detection and location to be partitioned into human eye topography.
In specific implementation, such as, Ada-Boosting (self adaptation strengthens study) the human eye detection localization method based on haar-like (class Ha Er) feature can be adopted to obtain position of human eye, and then be partitioned into human eye topography.It should be noted that, Haar-like feature, namely the Haar feature often said of existing technical staff, is a kind of conventional feature interpretation operator of computer vision field.Haar-like feature conventional at present can be divided three classes: linear character, edge feature, point patterns (central feature), diagonal feature.
Step S109: judge whether the image quality parameter that the human eye topography obtained has is positioned at default eye image mass parameter number range (namely carrying out eye image quality evaluation), if, then described human eye topography is stored in real time, if not, execution step S107 is returned;
For the present invention, in step S109, described human eye topography can be stored into cloud platform, remote server or other devices with data storage function.
Step S110: the iris region in human eye topography is positioned, then extract face characteristic that the facial image of active user that stores has and extract the iris feature that in human eye topography, iris region has, and the face characteristic that has respectively of the face characteristic this user had and iris feature and the multiple validated users (as registered users) prestored and iris feature information compare and mates, complete identifying.
In the present invention, it should be noted that, in specific implementation, in step s 110, specifically can adopt iris locating method, rubbersheetmodel (rubber slab model) algorithm of such as Daugman (name), positions the iris region in human eye topography.
In the present invention, it should be noted that, existing Feature Extraction Technology can be utilized to extract face characteristic that the facial image of active user has and extract the iris feature that in human eye topography, iris region has.Feature Extraction Technology is the basic conception of image procossing and area of pattern recognition, and the present invention does not illustrate know-why at this.Feature Extraction Technology can select suitable feature according to image property, is beneficial to carry out match cognization.
For the present invention, in step s 110, described comparison match step is specially: if the face characteristic that has of the face characteristic that has of this user and iris feature and one of them validated user and iris feature information match, then judge that active user is as validated user, if the face characteristic do not had with any one validated user and iris feature information match, then judge that active user is as disabled user, completes identifying.
For the present invention, in step s 110, it should be noted that, the face characteristic that the described multiple validated users (as registered users) prestored have respectively and iris feature information, step 101 of the present invention can be run respectively to multiple validated user to obtain to step S109, by to multiple validated user registered in advance, and store the face characteristic and iris feature information that they have.
Therefore, known based on above technical scheme, the remote multi-modal biological characteristic recognition methods of one provided by the invention, utilize face mode and the advantage of iris mode in living things feature recognition, by building active light source and polynary camera imaging system, merge intelligent human-machine interaction mode and biological characteristic detection method, breach existing living creature characteristic recognition system and carrying out the technical bottleneck in remote living things feature recognition, to promotion living things feature recognition applying on a large scale in scene, there is important practical significance.
See Fig. 2, in specific implementation, in order to run the remote multi-modal biological characteristic recognition methods of one that the invention described above provides, present invention also offers a kind of remote multi-modal biological characteristic recognition system, this system comprises:
Depth camera 101, for the depth image of shooting, collecting user, obtains the depth information (three-dimensional feature information of user) of user;
For the present invention, in specific implementation, can by the depth image of depth camera shooting, collecting user, and export the depth distance of user position, particular by multi views anaglyph, photometric stereo, colourity forming process or defocus the existing depth computing methods based on image such as deduction method and obtain depth distance between user position and spot for photography (as depth camera infield).
For the present invention, in specific implementation, described depth camera 101 can select in TOF (time-of-flight method) video camera, structured light depth camera and laser scanning depth camera any one.
Central processing control module 102, be connected with depth camera 101, for controlling depth video camera 101 startup optimization, according to the depth information of the user that depth camera obtains, obtain the depth distance (distance namely between user position and spot for photography) of user position, and judge whether the depth distance of user position is positioned within the distance range of user preset, if, then send default facial image shooting to user and start information, and trigger operation face video camera 103, if not, when described depth distance is less than the minimum threshold values of user preset distance range or when being greater than maximum threshold values, correspondence sends default facial image shooting and prepares information respectively,
For the present invention, in specific implementation, described central processing control module 102 can send default facial image shooting by loud speaker and/or display to user and start information.
For the present invention, in specific implementation, described central processing control module 102 can be connected with user instruction input equipments (human-computer interaction module) such as keyboard, mouse, touch-screen, tablets, user not only can be facilitated to input identification sign on, thus control central processing control module 102, user can also be facilitated to pre-set present system, such as, pre-set distance range, the number range that quality of human face image parameter is set etc.
In the present invention, in specific implementation, described central processing control module by be connected with the output equipment such as loud speaker, display screen (human-computer interaction module), can send default facial image shooting to user and starts information.
For the present invention, for described central processing control module 102, in specific implementation, the shooting of described facial image start information can according to user need arrange in advance, can be such as one section of sound of " movement please be stop, standing still and look squarely front " for the content of played pre-recorded.
For the present invention, for described central processing control module 102, in specific implementation, when described depth distance is less than the minimum threshold values of user preset distance range, the default described facial image shooting that correspondence sends prepares one section of sound that information can be " please moving backward " for the content prerecorded; When described depth distance is greater than the maximum threshold values of user preset distance range, the facial image the preset shooting that correspondence sends prepares one section of sound that information can be " please move forward " for the content prerecorded.
Face video camera 103, for the direct picture of shooting, collecting user;
For the present invention, can be carried out the direct picture of shooting, collecting user by face video camera, described face video camera can adopt the video camera of COMS or ccd imaging sensor.
In the present invention, in specific implementation, described face video camera 103 is preferably and adopts autozoom video camera or can realize digital lightfield camera of heavily focusing.
Central processing control module 102, also be connected with face video camera 103, for carrying out Face datection to the direct picture of the captured user gathered of face video camera 103, then locating segmentation obtains the facial image in user's direct picture, and judge whether image quality parameter that the facial image that obtains has is positioned at the number range (namely carrying out quality of human face image evaluation) of default quality of human face image parameter, if, then described facial image real-time storage is stored recognition unit 106 to preset data, and trigger operation light source module 104 and iris video camera 105, if not, continue the direct picture controlling face video camera shooting, collecting user,
For the present invention, it should be noted that, Face datection and location belong to specific algorithm part, in specific implementation, the result of the Face datection of existing maturation and location scheduling algorithm can be utilized to be partitioned into facial image.
In specific implementation, such as can adopt and carry out Face datection based on haar (Ha Er) classifier algorithm, utilize SDM (SupervisedDescentMethod) (having the gradient descent method of supervision) algorithm to carry out face key point location.
For the present invention, described image quality parameter can comprise in resolution, contrast and colour temperature one or more.The number range of described default quality of human face image parameter, can according to user need arrange in advance.
For the present invention, described preset data stores recognition unit can for having the cloud platform of data storage function, remote server or other devices.
Light source module 104, for sending the near infrared light needed for iris imaging to user;
Iris video camera 105, for the face subregion image of shooting, collecting user, and obtains human eye area image (namely laterally covering two face with eyebrow, the bridge of the nose at interior one section of image) from described face subregion Iamge Segmentation;
It should be noted that, for the present invention, iris video camera 105, the facial image testing result that can obtain according to central processing control module 102 and face video camera, nominal data between iris video camera, can take to dig window shape formula from the image of face subregion, split acquisition human eye area image, digging window is exactly only interested region (referring to human eye area) in the area image of protoplast face part is split.
Described nominal data refers to the pixel corresponding relation that in different cameras image, overlapped fov is interregional, the human eye key point position in facial image can be obtained by central processing control module 102, according to the demarcation relation of face video camera and iris video camera, and then the relevant position of the corresponding human eye key point of face subregion image can be obtained, can by human eye area Iamge Segmentation out according to this position.
For the present invention, near infrared light needed for iris imaging can be sent by the light source that is made up of LED array to user.
For the present invention, can by the face subregion image of existing iris video camera shooting, collecting user.
In the present invention, in specific implementation, described iris video camera is preferably and adopts autozoom video camera or can realize digital lightfield camera of heavily focusing.
Central processing control module 102, also be connected with iris video camera 105, for controlling iris video camera 105 startup optimization, and to iris video camera 105 gather the human eye area image of user obtained and carry out human eye detection, then locating segmentation obtains the human eye topography in user's human eye area image, and judge whether the image quality parameter that the human eye topography obtained has is positioned at default eye image mass parameter number range (namely carrying out eye image quality evaluation), if, then described human eye topography real-time storage is stored recognition unit 106 to preset data, if not, continue controlling run light source module 104 and iris video camera 105,
For the present invention, it should be noted that, human eye detection and location belong to specific algorithm part, and in specific implementation, the present invention can utilize the result of this kind of algorithms such as existing human eye detection and location to be partitioned into human eye topography.
In specific implementation, such as, Ada-Boosting (self adaptation strengthens study) the human eye detection localization method based on haar-like (class Ha Er) feature can be adopted to obtain position of human eye, and then be partitioned into human eye topography.It should be noted that, Haar-like feature, namely the Haar feature often said of existing technical staff, is a kind of conventional feature interpretation operator of computer vision field.Haar-like feature conventional at present can be divided three classes: linear character, edge feature, point patterns (central feature), diagonal feature.Preset data stores recognition unit 106, be connected with described central processing control module 102, the user's facial image sent for central processing control module described in real-time storage 102 and human eye topography, and the iris region in human eye topography is positioned, extract face characteristic that the facial image of active user that stores has and extract the iris feature that in human eye topography, iris region has, and the face characteristic that has respectively of the face characteristic this user to be had and iris feature and the multiple validated users (as registered users) prestored and iris feature information compare and mate, complete identifying.
In the present invention, it should be noted that, in specific implementation, described preset data stores recognition unit specifically can run iris locating method, rubbersheetmodel (rubber slab model) algorithm of such as Daugman (name), positions the iris region in human eye topography.
The present invention, described preset data stores recognition unit can for having the cloud platform of data storage function, remote server or other devices.
In the present invention, it should be noted that, the iris feature that in existing Feature Extraction Technology can be utilized to extract face characteristic that the facial image of active user has and human eye topography, iris region has.Feature Extraction Technology is the basic conception of image procossing and area of pattern recognition, and the present invention does not illustrate know-why at this.Feature Extraction Technology can select suitable feature according to image property, is beneficial to carry out match cognization.
For the present invention, the face characteristic had respectively in order to the face characteristic that this user had and iris feature and the multiple validated users (as registered users) prestored and iris feature information compare mates, the concrete processing procedure that described preset data stores recognition unit 106 is: if the face characteristic that has of the face characteristic had for this user and iris feature and one of them validated user and iris feature information match, then judge that active user is as validated user, if the face characteristic do not had with any one validated user and iris feature information match, then judge that active user is as disabled user, complete identifying.
For the present invention, it should be noted that, the face characteristic that the described multiple validated users (as registered users) prestored have respectively and iris feature information, can according to technical scheme of the present invention, multiple validated user is run in advance respectively to the acquisitions such as depth camera 101 of the present invention, central processing control module 102, face video camera 103, light source module 104, iris video camera 105, can by multiple validated user registered in advance, and store the face characteristic and iris feature information that they have.
For the present invention, in specific implementation, described central processing control module 102 can comprise multiple central processor CPU or digital signal processor DSP.
For the remote multi-modal biological characteristic recognition system of one provided by the invention, it can identify the distance range of user, can be 1 meter to 3 meters in the ordinary course of things.
Fig. 3 is the schematic appearance of a kind of remote multi-modal biological characteristic recognition system provided by the invention.In specific implementation, see Fig. 3, described remote multi-modal biological characteristic recognition system comprises system shell 100, the lower end, front of described system shell 110 has central processing control module 102, light source module 104 is provided with directly over described central processing control module 102, display screen 200 is provided with directly over described light source module 104, the both sides up and down of described display screen 200 are respectively arranged with an iris video camera 105, described two iris video cameras 105 are in the vertical direction in distributing up and down, with the collection of the face subregion image of the user of satisfied different height.
It should be noted that, for remote multi-modal biological characteristic recognition system provided by the invention, if comprise vertical direction in two the iris video cameras 105 distributed up and down, so described central processing control module 102 is also for carrying out Face datection by the direct picture of the user to the captured collection of face video camera 103, judge the height of user, and according to the numerical values recited of height, corresponding unlatching is positioned at above display screen or the iris video camera 105 of below.
For the present invention, in specific implementation, according to locating human face's image-region on the original direct picture that Face datection result can be taken at face video camera, according to the present position of facial image region on original direct picture, by user's height and the facial image region corresponding relation on original direct picture between present position pre-set, the height obtaining user can be judged.
The left side of described display screen 200 is provided with face video camera 103, and the right side of described display screen 200 is provided with depth camera 101, and the top of described depth camera 101 has loud speaker 300.
It should be noted that, in the present invention, the field range of face video camera 103 and depth camera 101 is larger than iris video camera 105, its position is not limited to shown in figure, only need ensure regardless of user's height and present position, face video camera 103 and depth camera 101 all can photograph user position interested.
In the present invention, see Fig. 3, described display screen 200, for carrying out image instruction to the position of user and last recognition result of the present invention, also can adopt the one that the mode of touch display screen inputs as system;
In the present invention, see Fig. 3, described loud speaker 300 is for making voice instruction to the position of user and recognition result.
In the present invention, in specific implementation, described face video camera 103, iris video camera 105, depth camera 101 can be connected with central processing control module 102 by any one connected modes such as USB3.0 or CameraLink respectively.
See Fig. 4 to Fig. 6, when described face video camera 103 and iris video camera 105 adopt autozoom video camera, autozoom mode has two kinds, a kind of is realize zoom (as shown in Figure 5) by the camera lens photocentre of mobile autozoom video camera, and another is moved by the transducer of control autozoom video camera to realize zoom (as shown in Figure 6).
It should be noted that, in Fig. 4 to Fig. 6,501 is subject, and 502 is the photocentre of the camera lens of autozoom video camera, and 503 is camera sensor, and 504 is transducer imaging.Autozoom video camera when Fig. 4 is non-zoom carries out imaging to being positioned at subject 501 at a distance; When changing camera lens photocentre 502 present position of autozoom video camera, as shown in Figure 5, the camera lens photocentre 502 of autozoom video camera moves forward, subject 501 and camera sensor 503 not movement when, the focal distance f of camera lens becomes large, view angle theta diminishes, then the picture on camera sensor 503 being as shown in Figure 5, achieves zoom function.When changing the position residing for camera sensor 503, as shown in Figure 6, move after camera sensor 503, the camera lens photocentre 502 of subject 501 and autozoom video camera not movement when, lens focus f becomes large, and view angle theta diminishes, and the picture that camera sensor 503 is as shown in Figure 6, consistent with Fig. 5 imaging, achieve zoom function equally.
For the present invention, as mentioned above, in specific implementation, described face video camera and iris video camera are preferably and adopt autozoom video camera or can realize digital lightfield camera of heavily focusing.
For described central processing control module 102, in order to control face video camera 103 and iris video camera 105 startup optimization, gathering respectively and obtaining the face of user or the shape library image of iris.Described central processing control module 102 comprises autozoom and controls submodule and numeral and heavily to focus control submodule, wherein:
Autozoom controls submodule, be connected with iris video camera with face video camera respectively, for when described face video camera and iris video camera adopt autozoom video camera, according to the depth distance value of the user present position that depth camera obtains, determine the focusing position of described face video camera and iris video camera, control described face video camera and iris camera driver and change the camera lens of video camera and the relative position of transducer, realize auto-focusing, obtain the shape library image of face or iris.
For the present invention, it should be noted that, in specific implementation, according to the image of depth camera shooting, can obtain the distance depth value residing for user, namely this distance value is the position that face and iris video camera need focusing.
For the present invention, also it should be noted that, in specific implementation, can pass through the distance between video camera internal mechanical structural change camera lens photocentre and transducer such as face video camera and iris video camera, this relative distance is exactly the focal length of camera lens, if the object distance of shooting is far away, then increase focal length, both increases relative distance, if the object distance of shooting is nearer, then reduce focal length, both reductions relative distance.
Numeral is heavily focused control submodule, be connected with iris video camera with face video camera respectively, for when described face video camera and iris video camera adopt lightfield camera, control described face video camera and the original face of iris camera acquisition and eye image and transfer to central processing control module, then, according to the depth distance of the user present position that depth camera obtains, determine the focusing position of described face video camera and iris video camera, then parameter of heavily focusing is determined, the heavy focus operation of numeral is carried out to original face and face subregion image, obtain the shape library image of face or iris.
For the present invention, it should be noted that, in specific implementation, according to the image of depth camera shooting, can obtain the distance depth value residing for user, namely this distance value is the position that face and iris video camera need focusing.
It should be noted that, heavily to focus parameter, this is the parameter that optical field imaging numeral is heavily focused, if the distance between the target of determining and camera, so this distance just corresponding parameter of heavily focusing, utilize this parameter, just can be obtained the shape library image of target location by the heavy focus algorithm of numeral.The heavy focus algorithm of numeral also a kind of known technology at last at present, does not specifically introduce here.
Therefore, known based on above technical scheme, the remote multi-modal biological characteristic recognition system of one provided by the invention, utilize face mode and the advantage of iris mode in living things feature recognition, by building active light source and polynary camera imaging system, merge intelligent human-machine interaction mode and biological characteristic detection method, breach existing living creature characteristic recognition system and carrying out the technical bottleneck in remote living things feature recognition, to promotion living things feature recognition applying on a large scale in scene, there is important practical significance.
In sum, compared with prior art, the invention provides a kind of remote multi-modal biological characteristic recognition methods and system thereof, its multi-modal biological characteristic recognition technology utilizing iris mode and face mode to combine, build high-precision optical system platform, adopt intelligent human-machine interaction mode, more traditional single living things feature recognition, not only can realize maximizing favourable factors and minimizing unfavourable ones between each feature, have complementary advantages, and can while remote identification, the further accuracy improving identification, realize remote, high-resolution, multi-modal biological characteristic identification at a high speed, the product use sense being conducive to improving user is subject to, be of great practical significance.
The above is only the preferred embodiment of the present invention; it should be pointed out that for those skilled in the art, under the premise without departing from the principles of the invention; can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.