A kind of anti-fraud face identification system based on 3-D scanningTechnical field
The invention belongs to technical field of mapping is and in particular to a kind of anti-fraud face identification system based on 3-D scanning.
Background technology
Face identification system, with face recognition technology as core, is an emerging biological identification technology, is the current worldThe high-quality precision and sophisticated technology of sciemtifec and technical sphere tackling key problem, its widely used regional characteristics analysis algorithm, has merged computer image processing technologyWith biostatisticss' principle in one, extract portrait characteristic point using computer image processing technology from video, using biologyStatistical principle is analyzed founding mathematical models, has vast potential for future development.
The principle of face identification system for system input be usually one or a series of containing the face not determining identityThe facial image recognition of some known identities in image, and face database or corresponding encode, and its output is then one to beRow similarity score, shows the identity of face to be identified.
Face identification system is widely used in recognition of face access management system, recognition of face access control and attendance system, faceIdentification monitoring management, recognition of face computer security are taken precautions against, recognition of face photo is searched for, recognition of face is come to visit registration, recognition of faceATM intelligent video warning system, recognition of face prison intelligent alarm system, recognition of face RFID intelligence clearance system, faceIdentification public security criminal pursues and captures an escaped prisoner intelligent alarm system etc. it is adaptable to park, factory, supermarket, cell square, conference centre, physical cultureThe public activities such as venue, school, hospital, residential quarter, commercial street, the large-scale market of farm produce and the significant points of gathering place, hotel（Hotel）, food and drink, public place of entertainment, the hall gateway of office building, multiple occasions such as elevator and other main thoroughfares.
Present face identification system is to be analyzed based on two-dimentional image, including man face image acquiring, Face detection,People is separated in a frame of a sub-picture or video flowing from background by the technology such as recognition of face contrast identity validation, willThe portrait capturing is contrasted with a certain object of registration in data base；The face of the set feature and PCA that are typically based on face is knownNot, after a series of training algorithms lifting precision, realized based on gradation of image.
But, the 3D model that the facial image acquisition system of two dimension is easily generated by photo is cheated, by openPhoto build dummy model can cheat most face identification system it is only necessary to three low come the mode to crack face detectionThe photo of resolution is with regard to enough to reappear 3D model that authentication system is cheated, fundamentally being referred to based on the fraudulence of VRTo the leak based on two-dimentional identification system, in the case that photographic quality is good enough, VR builds the two dimension that can break through more than 80%Face identification system.
Content of the invention
Not enough for prior art, the invention provides a kind of anti-fraud face identification system based on 3-D scanning, shouldSystem has that structure is simple, reasonable in design, efficient, safe and reliable advantage.
For realizing goal of the invention, the present invention employs the following technical solutions：
A kind of anti-fraud face identification system based on 3-D scanning, obtains system, three-dimensional information storage system including three-dimensional informationSystem and three-dimensional information comparison system, described three-dimensional information storage system obtains system, three-dimensional information contrast respectively with three-dimensional informationSystem connects.
Be preferably, described three-dimensional information obtain system include powering module and connected projector module, video camera,Face tracking and depth transducer, light filling module and communication module group, are provided with projector ray machine in described projector module.
Be preferably, described face tracking and depth transducer include control unit and respectively connected image unit,Face tracking module, sensing unit, display unit, driver element, described driver element connecting power device.
Be preferably, described three-dimensional information storage system include integrated edge extracting module on circuit boards, noise reduction module,Eliminate expression weight module, texture removes splicing and resets module, three-dimensional feature quantifies module and time storage module.
The present invention uses depth to pass in the moment of collection by optical camera and projector ray machine come gathered data informationSensor to be built three-dimensional coordinate system and to obtain texture model with high-definition camera camera lens, then the data of acquisition is passed toRebuild in information system and analyzed, repeated to shoot multiple image, then go to calculate the locus of face by three-dimensional algorithm（Coordinate is constructed with depth transducer and optical lens in scanning process）, can be spliced in real time afterwards, be obtainedComplete three-dimensional data.
It is preferably, described three-dimensional information obtains system and includes power module and connected laser instrument module, HDR industryCamera, color cmos camera, light filling module and communication module group, are provided with laser instrument in described laser instrument module.
It is preferably, described three-dimensional information storage system includes integrated discrete point cloud on circuit boards and processes module, secondarySurface interpolation module, edge extracting module, noise reduction module, eliminate expression weight module, texture remove splicing with reset module, threeDimensional feature quantifies module and time storage module.
The present invention gathers number by high resolution camera, laser instrument module, red flashing light lamp and two HDR industrial camerasIt is believed that breath, laser rays are irradiated to human face, two cameras catching this flashy 3 d scan data, due to people's body surfaceThe curvature in face is different, and light is radiated on object and can occur to reflect and reflect, and now scanner will obtain the space of object under testInformation, depth information, for determining each point position in space of body surface, 3D scanner, in different angles, can be rememberedSome width positional informationes of record object, these positional informationes are compared by software（align）, to obtain overall positional information, softPart merges（fuse）All of positional information is combined into a complete object, at this moment can be changed by these information softwaresFor 3D point cloud, and cloud data only comprises geometric coordinate and the Reflection intensity information of object, by carrying out inside and outside ginseng to cameraNumber demarcate after, by photogrammetric mode by image and point cloud merged, collection moment use depth transducer withIndustrial camera camera lens, to build three-dimensional coordinate system, obtains texture mapping data with the principle that a cloud spreads again, then will obtainThe data taking is passed in information system to be rebuild and is analyzed, and repeats to shoot multiple image, then goes to calculate by three-dimensional algorithmThe locus of face（Locus are determined with laser sensor in scanning process）, can be spelled in real time afterwardsConnect, obtain complete three-dimensional data.
It is preferably, described three-dimensional information obtains system and includes power module and connected optical camera, RGB colorCamera, 3D depth induction apparatuss, two dimensional character tracker module, light filling module and communication module group, described 3D depth induction apparatuss bagInclude RF transmitter and infrared C MOS camera.
Be preferably, described three-dimensional information storage system include integrated demarcation module on circuit boards, edge extracting module,Noise reduction module, elimination expression weight module, texture remove splicing and quantify module and time storage mould with replacement module, three-dimensional featureGroup.
It is preferably, described three-dimensional information comparison system includes the main control server of 1 internal database, right for informationThan, feedback, preserve.
It is preferably, Raltiva chip that described optical camera includes being sequentially connected in order, autozoom module, focal lengthAnd aperture.
It is preferably, described light filling module includes circuit board, metal rack, pad and light compensating lamp, described metal rack two endsConnect circuit board and pad respectively, described light compensating lamp is fixed in pad.
The present invention passes through optical camera, RGB color camera, RF transmitter and infrared C MOS camera institute structureThe 3D depth induction apparatuss becoming carry out gathered data information, and 3D depth induction apparatuss to perceive environment by way of black and white spectrum：BlackRepresent infinity, pure white represent infinitely near, the corresponding object of gray zone between black and white arrives the physical distance of sensor, and its collection regardsEvery bit in wild scope, and form the depth image that a width represents surrounding, the depth of field is generated with the speed of 10-30 frame per secondImage stream, real-time 3D ground reproduces surrounding, and software can carry out Pixel-level assessment to depth image, to distinguish the different portions of human bodyPosition, meanwhile, this process must shorten response time with the pretreatment optimizing, and is based on flight time TOF in the moment of collectionThe principle of the depth calculation imaging of camera, collocation depth transducer and high definition camera camera lens come to build three-dimensional coordinate system withObtain texture model, then the data of acquisition is passed in information system is rebuild and analyze, regard with acquired multiframeFrequency flows, then goes to calculate the locus of face by three-dimensional algorithm（Depth transducer and optical frames have been used in scanning processHead constructs coordinate）, can be spliced in real time afterwards, obtain complete three-dimensional data, reused ActiveApperance Model as two dimensional character tracker, then the depth data of calculated model extension to the systemOn, carry out three-dimensional face identification.
Human body will be distinguished from background environment by 3D depth induction apparatuss using segmentation strategy, extract from noiseGo out useful signal.Can be with face's skeleton of Proactive traceback user, in this stage, we are each tracked object in the depth of fieldSo-called segmentation shade is created, this is a kind of by background object in image（Such as chair and house pet etc.）Depth map after rejectingPicture, in handling process below, only the part of shade is split in transmission, to mitigate body-sensing amount of calculation, obtains the image of multiframeAfterwards, next just it is by the work recognizing, the simply basic image data being obtained through LightCoding technology, weightPoint is also intended to recognize image, is converted to effective instruction.
The invention has the beneficial effects as follows：The invention provides a kind of anti-fraud face identification system based on 3-D scanning,This system has that structure is simple, reasonable in design, efficient, safe and reliable advantage.
1st, for respect to being identified to face with plane geometry principle, the system uses space geometry principle, leads toCross the acquisition of information of depth dimension, make the structure of VR model recognition of face will not be caused with fraud it is necessary to be the three-dimensional in spaceObject could be by the checking of recognition of face；
2nd, for the face identification method of feature based PCA/LDA, the system does not need to carry out the change of image, butThe threedimensional model being directed in data base is compared, and success rate is more than more than 80%；
3rd, the system can obtain texture mapping, exclusion illumination, skin color, after human face expression impact, can by eyeball, scar,A series of indexs such as cheekbone height are added to recognition of face, and on the basis of anti-fraud, 3-D scanning would is that springformThe face coupling of type, judges the true and false by the information of multiple dimensions.
Brief description
Fig. 1 is the structured flowchart of the present invention；
Fig. 2 is the structural representation that in embodiment 1, three-dimensional information obtains system；
Fig. 3 is the structural representation that in embodiment 2, three-dimensional information obtains system；
Fig. 4 is the structural representation that in embodiment 3, three-dimensional information obtains system；
Fig. 5 is the structural representation of three-dimensional information storage system of the present invention；
Fig. 6 is the structural representation of three-dimensional information comparison system of the present invention；
Fig. 7 is the structured flowchart of face tracking of the present invention and depth transducer；
Fig. 8 is the structural representation of light filling module of the present invention；
Fig. 9 is the structural representation of optical camera of the present invention；
In the figures above：1st, power supply module；2nd, projector module；3rd, face tracking and depth transducer；4th, light filling module；5th, lead toNews module；6th, video camera；7th, HDR industrial camera；8th, color cmos camera；9th, RGB color camera；10th, infrared C MOS takes the photographShadow machine；11st, circuit board；12nd, metal rack；13rd, pad；14th, light compensating lamp；15th, Raltiva chip；16th, autozoom module；17th, focal length；18th, aperture.
Specific embodiment
The present invention is further illustrated to pass through specific embodiment below in conjunction with the accompanying drawings.
Embodiment 1
As shown in Figure 1, Figure 2, shown in Fig. 5, Fig. 6, Fig. 7, Fig. 8, a kind of anti-fraud face identification system based on 3-D scanning, including threeDimension Information Acquisition System, three-dimensional information storage system and three-dimensional information comparison system, described three-dimensional information storage system respectively withThree-dimensional information obtains system, three-dimensional information comparison system connects, described three-dimensional information obtain system include powering module 1 and and itsProjector module 2, video camera, face tracking and depth transducer 3, light filling module 4 and the communication module group 5 connecting, described projectionIt is provided with projector ray machine, described face tracking and depth transducer 3 include control unit and difference is connected in instrument module 2Image unit, face tracking module, sensing unit, display unit, driver element, described driver element connecting power device, instituteState light filling module 4 and include circuit board 11, metal rack 12, pad 13 and light compensating lamp 14, described metal rack 12 two ends connect respectivelyConnect circuit board 11 and pad 13, described light compensating lamp 14 is fixed in pad 13, and described three-dimensional information storage system includes being integrated inEdge extracting module on circuit board, noise reduction module, elimination expression weight module, texture remove splicing and replacement module, three-dimensionalCharacteristic quantification module and time storage module, described three-dimensional information comparison system includes the main control server of 1 internal database,For the contrast of information, feedback, preserve.
Operation principle：
When using the system, user will carry out the recognition of face of three-dimensional before camera lens, and three-dimensional three-dimensional information obtains system firstUse light filling module, light filling is carried out to experiencer's face, to realize removing the effect of shade, by controlling projector module and shootingMachine obtains the three-dimensional light of user along data and texture mapping, carries out the acquisition of information of more than 3 times, face tracking and depth transducerObject will be tracked and obtain with the distance of curved surface of face part, with TCP/IP communication module group in face moving processAcquired data is uploaded to three-dimensional information storage system, face surface data carries out edge line extraction, object noise reduction, removeThe spatial warping that face brings because of expression, three-dimensional characteristic quantification removes texture and is spliced, reset afterwards,Texture mapping will be mapped in threedimensional model, can be locked for the time uploaded, such as no identify within 5min, willCarry out alarm feedback, main control server obtains the final data uploading, and enters with reserved three-dimensional face data in data baseRow contrast, after the contrast of multiple index such as face depth, cicatrix, cheekbone height, human eye, by feedback simultaneously to original dataLibrary file is modified.
Embodiment 2
As shown in Fig. 1, Fig. 3, Fig. 5, Fig. 6, Fig. 8, a kind of anti-fraud face identification system based on 3-D scanning, including three-dimensional letterBreath acquisition system, three-dimensional information storage system and three-dimensional information comparison system, described three-dimensional information storage system respectively with three-dimensionalInformation Acquisition System, three-dimensional information comparison system connect, and described three-dimensional information obtains system and includes powering module 1 and connectedLaser instrument module, HDR industrial camera 7, color cmos camera 8, light filling module 4 and communication module group 5, in described laser instrument moduleIt is provided with laser instrument, described light filling module includes circuit board 11, metal rack 12, pad 13 and light compensating lamp 14, described metal rack12 two ends connect circuit board 11 and pad 13 respectively, and described light compensating lamp 14 is fixed in pad 13, described three-dimensional information storage systemSystem includes integrated discrete point cloud on circuit boards and processes module, quadratic surface interpolation module, edge extracting module, noise reduction mouldGroup, elimination expression weight module, texture remove splicing and quantify module and time storage module, institute with replacement module, three-dimensional featureState the main control server that three-dimensional information comparison system includes 1 internal database, for the contrast of information, feedback, preserve.
Operation principle：
When using the system, user will carry out the recognition of face of three-dimensional before camera lens, and three-dimensional information obtains system first with mendingLight module, carries out light filling to experiencer's face, to realize removing the effect of shade, by controlling laser module（Built-in laserDevice）, 2 HDR industrial cameras obtain the three-dimensional space data of user, obtain texture mapping with the color cmos camera of low noise,By the spatial information of acquisition object under test, depth information, for determining body surface, each puts in space spatial digitizerPosition difference angle, can record some width positional informationes of object, these positional informationes are compared by software（align）, to obtainOverall positional information, software integration（fuse）All of positional information is combined into a complete object, at this moment can pass throughThese information softwares are converted to 3D point cloud, and cloud data only comprises geometric coordinate and the Reflection intensity information of object, lead toCross and camera is carried out after inside and outside parameter demarcation, by photogrammetric mode, image is merged with point cloud, carry out more than 3 timesAcquisition of information, the face tracking module of high frequency will be tracked to object in face moving process, and depth transducer obtainsTake the distance of curved surface of face part, by point cloud registering to controlling under net coordinate system, the point cloud for target disappearance is it is also possible to utilizePublic territory find same place it is carried out two-by-two registration, in order to by had a cloud be transformed under unified control net coordinate system withControl point registration method obtains a cloud and fits over together, requires wherein one station to be necessary for being fitted on control net coordinate system two-by-two during registrationUnder point cloud, the data of point cloud field data acquisition is directed into during software can be set up discrete in a large number according to coordinate points automatic MosaicSpatial point cloud data after, to construct the threedimensional model of human body on this basis, the leak reparation of point cloud is due to a cloud itselfDiscreteness, can lead to model to there is certain defect, need after the polygon stage carries out the operation such as auto-mending, adjustment to it,Carry out quadratic surface interpolation module and just can obtain accurate mathematical model, with TCP/IP communication module group by acquired dataPass to three-dimensional information storage system, face surface data carries out edge line extraction, object noise reduction, remove face and carry because of expressionThe spatial warping coming, three-dimensional characteristic quantification removes texture and is spliced, reset afterwards, a cloud is spread to three-dimensional againIn space, the pinup picture of required texture is mapped in three-dimensional point cloud, can be locked for the time uploaded, such as in 5minWithin no identify, will carry out alarm feedback, main control server obtains the final data uploading, reserved three with data baseDimension human face data is contrasted, after the contrast of multiple index such as face depth, cicatrix, cheekbone height, human eye, by feedback simultaneouslyOriginal database file is modified.
Embodiment 3
As shown in Fig. 1, Fig. 4, Fig. 5, Fig. 6, Fig. 8, Fig. 9, a kind of anti-fraud face identification system based on 3-D scanning, including threeDimension Information Acquisition System, three-dimensional information storage system and three-dimensional information comparison system, described three-dimensional information storage system respectively withThree-dimensional information obtains system, three-dimensional information comparison system connects, described three-dimensional information obtain system include powering module 1 and and itsThe optical camera that connects, RGB color camera 9,3D depth induction apparatuss, two dimensional character tracker module, light filling module 4 and logicalNews module 5, described 3D depth induction apparatuss include RF transmitter and infrared C MOS camera 10, described optical camera bagInclude the Raltiva chip 15 being sequentially connected in order, autozoom module 16, focal length 17 and aperture 18, described light filling module 4 wrapsInclude circuit board 11, metal rack 12, pad 13 and light compensating lamp 14, described metal rack 12 two ends connect circuit board 11 and weldering respectivelyDisk 13, described light compensating lamp 14 is fixed in pad 13, and described three-dimensional information storage system includes integrated demarcation on circuit boardsModule, edge extracting module, noise reduction module, elimination expression weight module, texture remove splicing and replacement module, three-dimensional feature amountChange module and time storage module, described three-dimensional information comparison system includes the main control server of 1 internal database, for believingThe contrast of breath, feedback, preservation.
Operation principle：
When using the system, user will carry out the recognition of face of three-dimensional before camera lens, and three-dimensional information obtains system first with mendingLight module carries out light filling to experiencer's face, to realize removing the effect of shade, to obtain high-quality with RGB color cameraProblem pinup picture, the 3D depth induction apparatuss that GB colour camera and RF transmitter and infrared C MOS camera are constituted withCollocation generates depth image stream with the speed of 10-30 frame per second, and real-time 3D ground reproduces surrounding, two dimensional character tracker moduleUsing segmentation strategy, face will be distinguished from background environment, from noise, extract useful signal, can be activelyFollow the trail of face's skeleton of user, in this stage, it is so-called that the system is that each tracked object creates in depth imageSegmentation shade, this is a kind of by background object（Such as chair and house pet etc.）Depth image after rejecting, in process belowIn flow process, only the part of shade is split in transmission, and to mitigate body-sensing amount of calculation, software can carry out Pixel-level assessment to depth image,To distinguish the face different parts of human body, meanwhile, this process must shorten response time with the pretreatment optimizing, in collectionThe depth calculation imaging based on flight time TOF camera for the moment principle, collocation 3D depth transducer and high definition camera camera lensTo build three-dimensional coordinate system and to obtain texture model, repeatedly to obtain video flowing, obtain face with 3D depth transducerPartial distance of curved surface employs Active Apperance Model as two dimensional character tracker, then calculatedTo our depth data, face tracking module will be tracked in face moving process model extension to object, obtainsAfter the image of multiframe, next just it is by the work recognizing, the simply basic shadow being obtained through LightCoding technologyAs data, emphasis is also intended to recognize image, is converted to effective instruction, obtains high-quality three-dimensional data, with TCP/IP communicationAcquired data is uploaded to three-dimensional information storage system by module, and face surface data carries out edge line extraction, object noise reduction,Remove the spatial warping that face brings because of expression, three-dimensional characteristic quantification removes texture and spliced, carries out weight afterwardsPut, texture mapping be mapped in threedimensional model, can be locked for the time uploaded, such as no identify within 5min,Alarm feedback will be carried out, main control server obtains the final data uploading, with reserved three-dimensional face data in data baseContrasted, after the contrast of multiple index such as face depth, cicatrix, cheekbone height, human eye, by feedback simultaneously to original numberIt is modified according to library file.
The foregoing is only the specific embodiment of the present invention, but the architectural feature of the present invention is not limited thereto, the present inventionCan be used on similar product, in the field of the invention, the change made or modification are all for any those skilled in the artCover among the scope of the claims of the present invention.